21 Bioinformatics and Computational Biology
21.1 When Biology Meets Computers
21.1.1 What Is Bioinformatics?
Bioinformatics = Using computers and math to understand biological data
Think of it like:
Using a calculator instead of counting by hand
Using GPS instead of paper maps
Using Google instead of encyclopedia
But for:
DNA sequences
Protein structures
Gene expression data
Evolutionary relationships
21.1.2 Why We Need Bioinformatics
The big data problem:
Human genome: 3 billion letters
Would take 95 years to read out loud!
Can’t analyze by hand
Need computers!
Modern biology generates MASSIVE data:
One sequencing run: Billions of DNA letters
Gene expression study: Thousands of genes
Proteomics experiment: Thousands of proteins
Too much for humans alone
Solution: Bioinformatics!
21.2 The Data Explosion
21.2.1 How Much Data?
Human Genome Project (1990-2003):
13 years to sequence one genome
3 billion base pairs
Cost: $3 billion
Today (2025):
Sequence genome in 1-2 days
Cost: <$1,000
Thousands sequenced per day!
Result: Exponentially growing data!
Genomic data is now BIGGER than:
YouTube
Twitter
Astronomy
We’re drowning in biological data - bioinformatics is the life raft!
21.3 Core Bioinformatics Tasks
21.3.1 1. Sequence Analysis
What it is: Analyzing DNA, RNA, or protein sequences
Common tasks:
Finding Genes:
Where are genes in a genome?
Start and stop codons
Splicing signals
ORF (Open Reading Frame) prediction
Sequence Alignment:
Comparing two or more sequences
Finding similarities and differences
Identifying conserved regions
Example:
Sequence 1: ATCGATCGATCG
Sequence 2: ATCG--CGATCG
**** ***** (matches)
**Motif Finding**:
- Searching for patterns
- Transcription factor binding sites
- Protein domains
- Regulatory elements
### 2. Genome Assembly
**The puzzle problem**:
- Sequencing breaks DNA into millions of pieces
- Like a jigsaw puzzle with billions of pieces!
- Computer assembles them back together
**How it works**:
1. Sequence millions of short DNA fragments
2. Find overlapping regions
3. Merge overlaps to build longer sequences
4. Repeat until whole genome assembled
**Challenges**:
- Repetitive sequences (same piece fits many places!)
- Sequencing errors
- Huge computational requirements
**Modern solutions**:
- Better algorithms
- Long-read sequencing helps
- Multiple technologies combined
### 3. Gene Annotation
**What it is**: Assigning biological meaning to raw genome sequences
Think of it this way:
- **Sequencing** gives you the letters (A, T, G, C)
- **Annotation** gives you the words and their meanings
- Like having a book in a foreign language vs. a translated, annotated edition!
#### Why Annotation Matters
**The problem**:
- Raw genome sequence = billions of letters
- Which parts are genes?
- What do those genes do?
- Where are regulatory elements?
- Without annotation, genome is just meaningless data!
**The solution**: Genome annotation!
**Real-world importance**:
- Disease gene identification
- Drug target discovery
- Understanding evolution
- Personalized medicine
- Agricultural improvements
#### Dynamic Nature of Annotation
**Annotations are constantly updated!**
**Human genome builds** (versions):
- **HG18** (2006) - NCBI Build 36
- **HG19** (2009) - GRCh37 - Used for ~10 years
- **HG38** (2013-present) - GRCh38 - Current standard
- Future builds coming as we learn more!
**What changes between builds?**
- Better sequence assembly
- New genes discovered
- Gene boundaries refined
- Regulatory elements added
- Errors corrected
**Amazing fact**: ~25% of human gene annotations have been modified in the last 2 years!
**Why so many updates?**
- New experimental data (RNA-seq, ChIP-seq, etc.)
- Better computational methods
- Improved understanding of biology
- Discovery of new regulatory mechanisms
**Practical implication**: Always note which genome build you're using in research!
#### Types of Annotation
There are **two main types** of annotation:
### Structural Annotation
**Definition**: Identifying the physical structure of genomic features
**What it identifies**:
**1. Protein-Coding Genes**:
- Gene boundaries (start and end)
- Exon locations
- Intron locations
- 5' UTR and 3' UTR
- Splice sites
- Start codon (ATG) and stop codons
**2. RNA Genes**:
- tRNA genes
- rRNA genes
- miRNA genes
- lncRNA genes
- snRNA, snoRNA genes
**3. Regulatory Regions**:
- Promoters (transcription start sites)
- Enhancers
- Silencers
- Transcription factor binding sites
- CpG islands
**4. Repetitive Elements**:
- **SINEs** (Short Interspersed Nuclear Elements)
- Alu elements in humans
- ~10% of human genome!
- **LINEs** (Long Interspersed Nuclear Elements)
- LINE-1 elements
- ~17% of human genome!
- **LTRs** (Long Terminal Repeats)
- From ancient retroviruses
- **Tandem repeats**
- Satellites, microsatellites
- **Transposons** ("jumping genes")
- DNA transposons
- Can move around genome!
**5. Other Features**:
- Pseudogenes (dead gene copies)
- Centromeres
- Telomeres
- Origins of replication
### Functional Annotation
**Definition**: Predicting the biological function of genomic features
**What it predicts**:
**1. Protein Function**:
- What does this protein do?
- What pathway is it in?
- What cellular process?
- Based on sequence similarity to known proteins
**2. Protein Domains**:
- Functional modules in proteins
- DNA-binding domains
- Kinase domains
- Membrane-spanning domains
- Use databases like Pfam, InterPro
**3. Gene Ontology (GO) Terms**:
- Standardized vocabulary for gene function
- Three categories:
- **Biological Process** (e.g., "DNA repair")
- **Molecular Function** (e.g., "ATP binding")
- **Cellular Component** (e.g., "nucleus")
- We'll explore GO in detail in the next section!
**4. Pathway Assignment**:
- Which metabolic pathway?
- Which signaling pathway?
- Use databases like KEGG, Reactome
**5. Disease Association**:
- Is gene linked to diseases?
- What mutations cause disease?
- Use databases like OMIM, ClinVar
**Example**:Structural annotation:
Gene: BRCA1
Location: Chromosome 17
Exons: 24
Length: 81,189 bp
Functional annotation:
Function: DNA repair, tumor suppressor
Domains: RING finger, BRCT domains
Pathway: Homologous recombination
Disease: Breast and ovarian cancer when mutated
GO terms: DNA repair, cell cycle checkpoint
#### Gene Prediction Approaches
How do computers find genes in raw DNA sequence? There are **three main approaches**:
### 1. Ab Initio Prediction (Pattern-Based)
**"Ab initio" means "from the beginning"** - no external reference needed!
**How it works**:
- Uses statistical models to recognize gene patterns
- Looks for gene "signals" in DNA sequence
- Purely computational - no comparison to other genomes
**What it looks for**:
**Gene Signals**:
- **Start codon**: ATG (marks where translation begins)
- **Stop codons**: TAA, TAG, TGA (marks where translation ends)
- **Splice sites**: GT...AG boundaries (intron-exon junctions)
- **Promoter elements**: TATA box, CAAT box
- **Poly-A signals**: AATAAA (marks mRNA end)
**Statistical Features**:
- **Open Reading Frames (ORFs)**: Long stretches without stop codons
- **Codon bias**: Organisms prefer certain codons (explained below!)
- **GC content**: Coding regions often have different GC% than non-coding
- **Periodicity**: Coding sequences show 3-base-pair periodicity
#### Understanding Codon Bias
**What is codon bias?**
**The genetic code is redundant**:
- 64 possible codons (4³ nucleotides)
- Only 20 amino acids + 3 stop codons
- Most amino acids have multiple codons (synonymous codons)
**Example - Leucine** has 6 codons:UUA, UUG, CUU, CUC, CUA, CUG → All code for Leucine
**Codon bias** = Non-random usage of synonymous codons
**Organisms prefer certain codons over others!**
**Example**:E. coli prefers:
- UUG for Leucine (used 50% of the time)
- UUA for Leucine (used 10% of the time)
Even though both code for same amino acid!
**Why codon bias exists?**
**1. tRNA Availability**:
- Cells have different amounts of each tRNA
- Preferred codons match abundant tRNAs
- Faster translation
- Higher expression
**2. Translation Efficiency**:
- Rare codons slow down translation
- Can cause ribosome stalling
- Affects protein folding
**3. mRNA Stability**:
- Certain codon patterns affect mRNA structure
- Influences half-life
**Codon Adaptation Index (CAI)**:
- Measure of codon bias (0 to 1)
- 1.0 = Uses only preferred codons
- 0.2 = Uses rare codons
- Highly expressed genes have CAI ~0.8-1.0
**Species-specific patterns**:
| Organism | GC Content | Preferred Codons |
|----------|------------|------------------|
| E. coli | ~50% | G/C at 3rd position |
| Yeast | ~40% | A/T at 3rd position |
| Human | ~40-45% | Variable by isochore |
| Plasmodium (malaria) | ~20% | Extremely A/T rich |
**Using codon bias in gene prediction**:
**Coding regions**:
- Show strong codon bias
- Match organism's preference
- Higher CAI score
**Non-coding regions**:
- Random codon usage
- No bias pattern
- Lower CAI score
**Ab initio gene finders use this**:Sequence with high CAI → Likely coding
Sequence with low CAI → Likely non-coding
**Practical application - Recombinant Protein Production**:
**Problem**: Express human gene in E. coli
- Human uses different codons than E. coli
- Gene has rare E. coli codons
- Low expression!
**Solution**: Codon optimization
- Replace human codons with E. coli preferred codons
- Same protein sequence
- 10-1000x higher expression!
**Example**:Human gene: ...UUA-CUA-UUA... (rare in E. coli)
Optimized: ...UUG-CUG-UUG... (preferred in E. coli)
Same protein, better expression!
#### ORF Detection Algorithm (Detailed)
**The computational method** to find ORFs:
**Step-by-step algorithm**:
**1. Consider all 6 reading frames**:
- **3 frames on forward strand** (+1, +2, +3)
- **3 frames on reverse complement** (-1, -2, -3)
**Example DNA sequence**:Forward: 5'-ATGCATGGATAA-3'
Reading frame +1: ATG CAT GGA TAA (start...stop) ← ORF!
Reading frame +2: A TGC ATG GAT AA
Reading frame +3: AT GCA TGG ATA A
Reverse complement: 3'-TACGTACCTATTT-5' = 5'-TTATAGGTACGTA-3'
Reading frame -1: TTA TAG GTA CGT A
Reading frame -2: T TAT AGG TAC GTA
Reading frame -3: TT ATA GGT ACG TA
**2. Scan for start codons (ATG)**:
- Mark all ATG positions in each frame
- Each is potential ORF start
**3. Extend until stop codon**:
- Read triplets from ATG
- Continue until TAA, TAG, or TGA
- This is one ORF
**4. Calculate ORF length**:
- Count nucleotides from start to stop
- Divide by 3 = number of codons
**5. Apply filters**:
- **Minimum length**: Usually >300 bp (100 codons)
- Too short → likely random
- **Codon bias**: Check if matches organism
- **Blast search**: Does it match known proteins?
**Example - Finding ORFs**:Sequence: 5'-ATGAAATTTGCATAA-3'
Frame +1: ATG AAA TTT GCA TAA
↑ ↑
Start Stop
ORF = ATG AAA TTT GCA TAA (15 bp, 5 codons)
Protein = M K F A * (4 amino acids)
Too short? If minimum is 100 codons, this is rejected!
**Longest ORF heuristic**:
- In a random sequence, expect short ORFs
- Real genes = long ORFs
- **Find longest ORF in each frame**
- Usually the real gene!
**Statistics**:
- Random sequence: Average ORF = 64 codons (192 bp)
- Real genes: Average = 300-500 codons (900-1500 bp)
**Prokaryotic vs. Eukaryotic ORF Finding**:
**Prokaryotes** (easier):
- No introns
- ORF = gene
- High gene density
- Start-to-stop is complete
**Eukaryotes** (harder):
- Introns interrupt ORFs!
- Need to predict exons
- Splice sites required
- Much more complex
**Tools for ORF Finding**:
- **ORF Finder** (NCBI): Web-based, simple
- **GeneMark**: Ab initio with species-specific models
- **Glimmer**: Especially for bacteria
- **Augustus**: For eukaryotes with introns
**Computational Methods**:
- **Hidden Markov Models (HMMs)**: Statistical models of gene structure
- **Neural networks**: Machine learning approaches
- **Support vector machines**: Classification algorithms
**Famous ab initio tools**:
- GeneMark
- Augustus
- GENSCAN
- GlimmerHMM
**Pros**:
✅ Works for novel genomes (no reference needed)
✅ Fast computation
✅ Can find unique genes
✅ Good for prokaryotic genomes (simpler gene structure)
**Cons**:
❌ Lower accuracy than homology-based methods
❌ Misses short exons
❌ Struggles with alternative splicing
❌ Can't detect overlapping genes well
❌ High false positive rate
**Best for**: First-pass annotation of new genomes, prokaryotic genomes
### 2. Homology-Based Prediction (Comparison-Based)
**"Homology" means similarity** - compares to known genes!
**How it works**:
- Compares new genome to databases of known genes
- Uses evolutionary conservation
- If sequence is similar to known gene, probably also a gene!
**The principle**:Known gene in mouse: ATGCCCAAAGGG
Your new sequence: ATGCCCAAGGGG
********** * (90% identical)
→ Probably same gene in your organism!
**Methods**:
**1. Protein-to-Genome Alignment**:
- Use known proteins from databases
- Align them to genome
- Tools: BLASTX, GeneWise, Exonerate
**2. Genome-to-Genome Alignment**:
- Compare entire genomes
- Human vs. chimpanzee (99% similar!)
- Human vs. mouse (useful for finding conserved genes)
- Tools: BLAST, BLAT
**3. Expressed Sequence Evidence**:
- Use mRNA/cDNA sequences
- Direct evidence of gene expression!
- RNA-seq data is gold standard
- EST (Expressed Sequence Tag) databases
**Example**:
- Sequenced new organism (e.g., bonobo)
- Compare to human genome (very similar!)
- Transfer annotations from human to bonobo
- Refine with experimental data
**Pros**:
✅ High accuracy for conserved genes
✅ Can transfer functional annotations
✅ Benefits from decades of research on model organisms
✅ Works well for eukaryotic genomes
**Cons**:
❌ Requires reference genome/proteins
❌ Misses species-specific genes
❌ Biased toward well-studied organisms
❌ Can miss rapidly evolving genes
**Best for**: Well-conserved genes, organisms with close relatives
### 3. Integrated Approach (Best of Both Worlds!)
**Combining multiple evidence sources** - the modern standard!
**What it integrates**:
1. **Ab initio predictions** (statistical signals)
2. **Homology evidence** (similarity to known genes)
3. **RNA-seq data** (experimental evidence of transcription)
4. **Protein data** (mass spectrometry evidence)
5. **Comparative genomics** (conservation across species)
6. **Epigenomic data** (chromatin marks, histone modifications)
**How it works**:Ab initio: Gene predicted at position X
Homology: Similar to mouse gene at position X
RNA-seq: Transcripts detected at position X
→ HIGH CONFIDENCE: Gene at position X!
Ab initio: Gene predicted at position Y
Homology: No match
RNA-seq: No expression
→ LOW CONFIDENCE: Probably false positive
**Modern integrated tools**:
- **MAKER**: Combines evidence, widely used
- **BRAKER**: Uses RNA-seq + ab initio
- **Ensembl pipeline**: Human genome annotation
- **NCBI pipeline**: RefSeq annotations
**Workflow**:
1. **Repeat masking** (remove repetitive DNA first!)
2. **Ab initio prediction** (first pass)
3. **Homology search** (compare to databases)
4. **RNA-seq alignment** (experimental evidence)
5. **Evidence integration** (combine all signals)
6. **Manual curation** (experts review tricky cases)
7. **Functional annotation** (predict gene functions)
**Pros**:
✅ **Highest accuracy**
✅ Catches genes missed by single methods
✅ Reduces false positives
✅ Provides confidence scores
✅ Standard for important genomes (human, mouse, etc.)
**Cons**:
❌ Requires multiple data types (expensive!)
❌ Computationally intensive
❌ Complex to implement
❌ Still not perfect - ~5-10% error rate
**Best for**: Important reference genomes, well-funded projects
## Gene Ontology (GO): A Universal Language for Gene Function
### What Is Gene Ontology?
**Gene Ontology (GO)** is a standardized classification system for describing gene and protein functions across all organisms [@geneontology2000; @geneontology2021].
Think of it like:
- 🌍 **Universal language** for biology (like how scientists worldwide use Latin names for species)
- 📚 **Dewey Decimal System** for genes (standardized classification)
- 🏷️ **Hashtags** for gene functions (standardized labels)
**Why we need GO**:
- Different organisms, same gene names mean different things
- Need consistent terminology across species
- Enables computational analysis
- Facilitates data sharing and integration
**Amazing fact**: GO terms are used to annotate genes from bacteria to humans!
### The Three Pillars of Gene Ontology
GO classifies genes into **three independent categories** (called ontologies):
**Figure 17.1**: The three main categories of Gene Ontology (GO): Molecular Function (what the molecule does), Biological Process (what larger process it's part of), and Cellular Component (where it's located).
*Image credit: Gene Ontology Consortium, CC BY 4.0*
### 1. Molecular Function (MF)
**Definition**: The **biochemical activity** of the gene product at the molecular level
Think of it as: **"What does this molecule DO?"**
**Examples**:
| GO Term ID | Term Name | What it Means | Example Genes |
|------------|-----------|-------------|---------------|
| GO:0003677 | DNA binding | Sticks to DNA | Transcription factors |
| GO:0004672 | Protein kinase activity | Adds phosphate groups to proteins | PKA, PKC |
| GO:0005524 | ATP binding | Grabs ATP molecules | Many enzymes |
| GO:0016491 | Oxidoreductase activity | Moves electrons between molecules | Cytochrome P450 |
| GO:0003723 | RNA binding | Sticks to RNA | RNA-binding proteins |
**Key points:**
- Describes the biochemical action
- Independent of where or when it happens
- Based on the inherent capability of the molecule
**Important**: A gene can have **multiple molecular functions**!
### 2. Biological Process (BP)
**Definition**: The **broader biological objective** or pathway that the gene product contributes to
Think of it as: **"What is this molecule PART OF?"**
**Examples**:
| GO Term ID | Term Name | What it Means | Example Genes |
|------------|-----------|-------------|---------------|
| GO:0006281 | DNA repair | Fixing damaged DNA | BRCA1, BRCA2 |
| GO:0007049 | Cell cycle | Cell division process | Cyclins, CDKs |
| GO:0006915 | Apoptosis | Programmed cell death | Caspases, p53 |
| GO:0006955 | Immune response | Fighting infections | Antibodies, cytokines |
| GO:0007165 | Signal transduction | Cellular signaling | Receptors, kinases |
**Key points:**
- Describes the larger biological context
- Multiple molecular functions contribute to one process
- Can be hierarchical (specific → general)
**Example hierarchy**:DNA metabolic process (broad)
└─ DNA repair (more specific)
└─ Double-strand break repair (very specific)
└─ Homologous recombination (most specific)
### 3. Cellular Component (CC)
**Definition**: The **location in the cell** where the gene product is active
Think of it as: **"WHERE in the cell is this molecule?"**
**Examples**:
| GO Term ID | Term Name | What it Means | Example Genes |
|------------|-----------|-------------|---------------|
| GO:0005634 | Nucleus | Inside nuclear membrane | Histones, transcription factors |
| GO:0005739 | Mitochondrion | Mitochondrial compartment | Respiratory chain proteins |
| GO:0005886 | Plasma membrane | Cell surface membrane | Receptors, ion channels |
| GO:0005737 | Cytoplasm | Cellular cytoplasm | Many metabolic enzymes |
| GO:0005576 | Extracellular region | Outside the cell | Secreted proteins, antibodies |
**Key points:**
- Describes subcellular localization
- Can include complexes (e.g., "ribosome")
- Important for understanding protein function
**Subcompartments** exist too:Mitochondrion (general)
├─ Mitochondrial matrix
├─ Mitochondrial inner membrane
└─ Mitochondrial intermembrane space
### GO Term Structure and Identifiers
**Every GO term has**:
- **Unique ID**: GO:XXXXXXX (7 digits)
- **Term name**: Human-readable name
- **Definition**: Precise description
- **Relationships**: Connections to other terms
**Example - GO:0003677 (DNA binding)**:ID: GO:0003677
Name: DNA binding
Ontology: Molecular Function
Definition: Any molecular function by which a gene product
interacts selectively and non-covalently with DNA
Synonyms: microtubule/chromatin interaction
Relationships:
- is_a: nucleic acid binding (GO:0003676)
- part_of: Various processes
**Hierarchical structure** (parent-child relationships):
- **More general terms** at top (parents)
- **More specific terms** below (children)
- Allows for different levels of detail
### Practical Example: Actin Gene
Let's annotate **Actin** (a cytoskeletal protein):
**Gene**: ACTB (β-actin gene)
**GO Annotations**:
**Molecular Function**:
- GO:0003779 - Actin binding
- GO:0005200 - Structural constituent of cytoskeleton
**Biological Process**:
- GO:0030036 - Actin cytoskeleton organization
- GO:0016477 - Cell migration
- GO:0051301 - Cell division
**Cellular Component**:
- GO:0005737 - Cytoplasm
- GO:0005856 - Cytoskeleton
- GO:0015629 - Actin cytoskeleton
**Why multiple terms?** Actin does many things in different contexts!
### Practical Example: BRCA1 Gene
**Gene**: BRCA1 (Breast Cancer 1, tumor suppressor)
**GO Annotations**:
**Molecular Function**:
- GO:0003677 - DNA binding
- GO:0004842 - Ubiquitin-protein transferase activity
- GO:0008270 - Zinc ion binding
**Biological Process**:
- GO:0006281 - DNA repair
- GO:0000724 - Double-strand break repair via homologous recombination
- GO:0007049 - Cell cycle
- GO:0006974 - Cellular response to DNA damage stimulus
- GO:0008283 - Cell proliferation
**Cellular Component**:
- GO:0005634 - Nucleus
- GO:0000785 - Chromatin
**Clinical relevance**: Mutations in BRCA1 disrupt DNA repair (GO:0006281), leading to cancer!
### How GO Annotations Are Created
**Two main approaches**:
**1. Manual Curation** (gold standard):
- Expert scientists read research papers
- Extract evidence for gene function
- Assign GO terms with evidence codes
- Time-consuming but accurate
**2. Computational Prediction**:
- Based on sequence similarity (homology)
- Transfer annotations from well-studied organisms
- Machine learning approaches
- Faster but less certain
**Quality codes:**
- **EXP**: Inferred from Experiment (most reliable)
- **IDA**: Inferred from Direct Assay
- **IMP**: Inferred from Mutant Phenotype
- **ISS**: Inferred from Sequence Similarity (computational)
- **IEA**: Inferred from Electronic Annotation (least reliable)
**Always check evidence codes!** Experimentally verified annotations are more trustworthy.
### Conservation Across Species
**GO enables cross-species comparisons!**
**Example - Actin**:
- Found in yeast, plants, mice, humans
- Same GO terms across species!
- GO:0003779 (Actin binding) applies to all
- Reveals evolutionary conservation
**Comparative genomics workflow**:
1. Identify homologous genes across species
2. Compare GO annotations
3. Conserved GO terms → essential functions
4. Species-specific GO terms → unique adaptations
### Using GO in Research
**Common applications**:
**1. Gene Set Enrichment Analysis**:
- Have list of 500 upregulated genes
- Which biological processes are enriched?
- Statistical test: Are certain GO terms over-represented?
- Reveals coordinated biological responses
**Example**:Input: 500 genes upregulated in cancer
Output:
- DNA repair (GO:0006281) - 50 genes (p < 0.001)
- Cell cycle (GO:0007049) - 75 genes (p < 0.001)
→ Cancer affects DNA repair and cell division!
**2. Functional Annotation of New Genomes**:
- Sequence unknown organism
- Find homologous genes with BLAST
- Transfer GO annotations from known organisms
- Predict functions computationally
**3. Disease Gene Discovery**:
- Identify disease-associated mutations
- Check GO annotations
- Understand disrupted pathways
- Design targeted therapies
**4. Drug Target Identification**:
- Need drugs for pathway X
- Query GO for genes in pathway X
- Prioritize druggable targets
- Accelerate drug discovery
### GO Databases and Tools
**Major resources**:
**1. Gene Ontology Consortium** (http://geneontology.org):
- Official GO database
- Browse/search GO terms
- Download annotations
- Documentation
**2. AmiGO** (Browser):
- User-friendly GO browser
- Search genes or GO terms
- Visualize hierarchies
- Export data
**3. QuickGO** (EBI):
- European GO browser
- Fast searching
- Annotation statistics
- API access
**4. GO Enrichment Tools**:
- **DAVID**: Functional annotation clustering
- **GOseq**: Accounts for gene length bias
- **topGO**: R package for enrichment
- **g:Profiler**: Web-based enrichment analysis
### Major Genome Annotation Repositories
**Where annotations are stored and accessed**:
**1. GENCODE** (https://www.gencodegenes.org):
- **Comprehensive gene annotation** for human and mouse
- Part of ENCODE project
- High-quality manual curation
- Includes:
- Protein-coding genes
- Long non-coding RNAs (lncRNAs)
- Small RNAs
- Pseudogenes
- **Current version**: Updates regularly
- **Human**: ~60,000 genes (20,000 protein-coding)
- Gold standard for human genome annotation
**What GENCODE provides**:
- Gene structures (exons, introns)
- Transcript isoforms (alternative splicing)
- Functional annotations
- GTF/GFF files for analysis
**2. RefSeq** (NCBI Reference Sequence):
- Curated non-redundant sequences
- Covers many organisms (not just human/mouse)
- Expert curation + computational methods
- Stable identifiers (e.g., NM_000546 for TP53)
- Used in clinical genetics
**3. Ensembl** (https://www.ensembl.org):
- European annotation resource
- Automated pipeline + manual curation
- Comparative genomics across species
- Gene trees and orthologs
- Variant annotation (Variant Effect Predictor)
**4. UniProt** (https://www.uniprot.org):
- For proteins specifically
- **Swiss-Prot**: Manually reviewed (gold standard)
- **TrEMBL**: Computationally annotated
- Functional information
- Protein domains, PTMs
- Literature references
**Comparison**:
| Database | Focus | Curation | Species Coverage |
|----------|-------|----------|------------------|
| **GENCODE** | Human/mouse genes | Manual + auto | Human, mouse |
| **RefSeq** | Reference sequences | Manual + auto | All organisms |
| **Ensembl** | Comparative genomics | Automated pipeline | 100+ vertebrates |
| **UniProt** | Proteins | Manual (Swiss-Prot) | All organisms |
**Version control is critical**:
- Annotations change over time
- Always cite which version used!
- Example: "GENCODE v44 (2023)"
**Accessing annotations**:
**Programmatic access**:
```bash
```bash
```bash
```bash
# Download GENCODE human annotations
wget ftp://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_human/release_44/gencode.v44.annotation.gtf.gz
# Download via Ensembl
wget ftp://ftp.ensembl.org/pub/release-110/gtf/homo_sapiens/
**Web browsers**:
- UCSC Genome Browser: Shows RefSeq, GENCODE, Ensembl tracks
- Ensembl browser: Shows Ensembl annotations
- Can compare different annotation sources side-by-side!
**Annotation file formats**:
**GTF (Gene Transfer Format)**:chr1 HAVANA gene 11869 14409 . + . gene_id "ENSG00000223972"
chr1 HAVANA transcript 11869 14409 . + . transcript_id "ENST00000456328"
chr1 HAVANA exon 11869 12227 . + . exon_id "ENSE00002234944"
**GFF3 (General Feature Format)**:
- Similar to GTF
- More flexible
- Supports hierarchical features
**Why multiple databases?**
**Different strengths**:
- **GENCODE**: Best for human/mouse RNA-seq analysis
- **RefSeq**: Best for clinical variant interpretation
- **Ensembl**: Best for comparative genomics
- **UniProt**: Best for protein function
**Discrepancies exist!**
- Same gene, different annotations
- Number of isoforms varies
- Exon boundaries differ slightly
- Choose database appropriate for your application
### Limitations of GO
**Important caveats**:
**1. Incompleteness**:
- Not all genes are annotated
- New functions continually discovered
- Bias toward well-studied organisms
**2. Annotation Quality Varies**:
- Some genes: extensive experimental evidence
- Others: only computational predictions
- Always check evidence codes!
**3. Context-Dependent Functions**:
- Gene may have different functions in different tissues
- GO doesn't always capture this nuance
- Temporal aspects not well represented
**4. Updating Annotations**:
- GO is constantly evolving
- Terms added, modified, made obsolete
- Need to use current version
- Citation matters!
### Best Practices for Using GO
**Do's**:
✅ Check evidence codes (prefer experimental)
✅ Use multiple GO terms (genes multitask!)
✅ Consider term hierarchies (specific → general)
✅ Keep GO database up to date
✅ Report GO version in publications
✅ Use enrichment analysis statistics properly
**Don'ts**:
❌ Don't rely solely on IEA (electronic) annotations
❌ Don't ignore term hierarchies
❌ Don't forget that annotations evolve
❌ Don't use GO terms in isolation
❌ Don't ignore p-values in enrichment analysis
### Future of Gene Ontology
**Ongoing developments**:
- **Causal Activity Models (CAMs)**: Representing molecular pathways
- **GO-CAM**: Integrating GO with pathway information
- **Better context representation**: Tissue-specific, temporal annotations
- **Machine learning**: Automated, high-quality predictions
- **Integration**: Linking GO with other ontologies (disease, anatomy, etc.)
**Vision**: Complete functional map of all genes in all organisms!
#### Challenges in Gene Prediction
**Why is gene prediction hard?**
**1. Alternative Splicing**:
- One gene → multiple transcripts
- Which exons are used when?
- ~95% of human genes alternatively spliced!
- Creates enormous complexity
**2. Overlapping Genes**:
- Genes on opposite DNA strands
- Genes within introns of other genes
- Hard for algorithms to detect
**3. Small Genes**:
- Short ORFs often missed
- Could be real genes or random
- MicroRNA genes are tiny!
**4. Pseudogenes**:
- Look like genes but are non-functional
- "Dead" copies from evolution
- ~20,000 pseudogenes in human genome
- Hard to distinguish from real genes
**5. Repetitive Sequences**:
- ~45% of human genome is repetitive!
- LINEs, SINEs, transposons everywhere
- Must be masked before annotation
- Can interfere with gene prediction
### 4. Variant Calling
**What it is**: Finding differences between genomes
**Why it matters**:
- Disease-causing mutations
- Drug response variants
- Evolutionary changes
- Personalized medicine
**Process**:
1. Sequence individual's genome
2. Compare to reference genome
3. Identify differences (SNPs, indels, etc.)
4. Annotate variants (what genes affected?)
5. Predict impact (harmful, benign, unknown?)
### 5. Gene Expression Analysis
**What it is**: Measuring which genes are active
**RNA-seq analysis**:
1. Sequence all RNA in sample
2. Count reads for each gene
3. Normalize data
4. Compare conditions (healthy vs. diseased)
5. Find differentially expressed genes
**Challenges**:
- Statistical analysis needed
- Multiple testing correction
- Biological vs. technical variation
### 6. Phylogenetic Analysis
**What it is**: Building evolutionary trees
**Process**:
1. Collect sequences from multiple species
2. Align sequences
3. Calculate evolutionary distances
4. Build tree showing relationships
**Uses**:
- Understanding evolution
- Tracking disease outbreaks (COVID-19!)
- Conservation biology
- Drug development
## Genome Browsers: Visualizing the Genome
### What Is a Genome Browser?
**Genome browser** = Interactive tool to visualize genome annotations and experimental data
Think of it like:
- Google Maps for the genome!
- Shows genes, regulatory elements, and experimental data
- Navigate by chromosome position
- Zoom in/out from whole chromosome to single nucleotide
**Why use genome browsers?**
- Visualize gene structure (exons, introns, UTRs)
- See regulatory elements (promoters, enhancers)
- View experimental data (RNA-seq, ChIP-seq, etc.)
- Compare across species
- Plan experiments
- Interpret variants
### UCSC Genome Browser (Most Popular!)
**UCSC** = University of California, Santa Cruz
**URL**: https://genome.ucsc.edu
**What it shows**:
**1. Main Display** (Genome View):Chromosome 17: 43,000,000 - 43,100,000 bp
RefSeq Genes: ▬▬▬▐██▐██▐██▐▬▬▬ BRCA1
└─┘ └┘ └┘ └┘
Introns│ Exons
Conservation: ████▒▒▒████▒▒▒██
(dark = conserved)
RNA-seq: ▁▁▃▆█▆▃▁▁▃▆█▆▃▁
(peaks = expression)
SNPs: | | | | | |
(genetic variants)
**2. Navigation Controls**:
- **Search by**: Gene name, position, keyword
- **Zoom**: In/out by 1.5x, 3x, 10x, 100x
- **Move**: Left/right along chromosome
- **Jump**: To specific region
### Key Tracks in UCSC Browser
**Tracks** = Layers of information displayed on the genome
**Gene Tracks**:
- **RefSeq Genes**: Curated gene annotations
- **GENCODE**: Comprehensive gene set
- **Shows**: Exons (thick boxes), introns (lines), UTRs (thin boxes)
- **Direction**: Arrow shows transcription direction
**Regulatory Tracks**:
- **CpG Islands**: Potential promoter regions
- **Transcription Factor ChIP-seq**: Where TFs bind
- **DNase Hypersensitivity**: Open chromatin regions
- **Enhancers**: Predicted regulatory elements
**Variation Tracks**:
- **dbSNP**: Single nucleotide polymorphisms
- **ClinVar**: Disease-associated variants
- **gnomAD**: Population allele frequencies
**Comparative Genomics**:
- **Conservation**: Evolutionary conservation across species
- **Vertebrate alignment**: Compare 100 vertebrate genomes
- **Shows conserved regions** (likely functional!)
**RNA-seq Tracks**:
- Gene expression levels
- Different tissues/conditions
- Alternative splicing visualization
**Epigenomic Tracks**:
- Histone modifications (H3K4me3, H3K27ac, etc.)
- DNA methylation
- Chromatin states (active, repressed, etc.)
### Using UCSC Browser: Example Workflow
**Goal**: Understand BRCA1 gene structure and regulation
**Steps**:
**1. Search**:
- Enter "BRCA1" in search box
- Browser zooms to BRCA1 on chromosome 17
**2. Gene Structure**:
- See 24 exons (thick blue boxes)
- See introns (lines connecting exons)
- See 5' and 3' UTRs (thin boxes)
- Total gene length: ~81 kb
- mRNA length: ~7.2 kb (much smaller - introns removed!)
**3. Regulatory Elements**:
- Enable "CpG Islands" track
- See CpG island at promoter (typical!)
- Enable "Layered H3K27Ac" (active enhancer mark)
- See several enhancers near gene
**4. Variants**:
- Enable "ClinVar" track
- See disease-causing mutations
- Red markers = pathogenic variants
- Click on variant for details (disease association, frequency, etc.)
**5. Conservation**:
- Enable "Conservation" track
- Dark peaks = highly conserved
- Conservation in exons (expected!)
- Also conservation in some introns (regulatory elements?)
**6. Expression**:
- Enable "GTEx Gene" track
- See BRCA1 expressed in many tissues
- Highest in testis, ovary, thymus
### Other Popular Genome Browsers
**Ensembl** (European)
- URL: https://www.ensembl.org
- Similar to UCSC
- Strong comparative genomics
- More international species
- Good variant effect predictor
**IGV (Integrative Genomics Viewer)**
- Desktop application (not web-based)
- Fast for large datasets
- Great for RNA-seq and variant viewing
- Researchers' favorite for detailed analysis!
**NCBI Genome Data Viewer**
- From National Center for Biotechnology Information
- Integrated with other NCBI resources
- Shows RefSeq annotations
### Practical Tips for Using Genome Browsers
**Understand coordinates**:
- **1-based**: UCSC uses 1-based coordinates
- **0-based**: BED files use 0-based
- Can cause off-by-one errors!
**Choose correct genome build**:
- HG38 (current)
- HG19 (older, still used)
- **Don't mix!** Coordinates differ between builds
**Custom tracks**:
- Upload your own data
- View alongside reference annotations
- Great for interpreting experiments
**Export images**:
- PDF or PNG for publications
- Customize colors and labels
**Session saving**:
- Save your track configuration
- Share with collaborators
- Reproducible views
### Real-World Use Cases
**1. Clinical Genetics**:
- Patient has variant in gene
- View in UCSC browser
- Check if variant in conserved region
- Check ClinVar for known disease association
- Predict pathogenicity
**2. Research**:
- Found interesting gene in RNA-seq
- View structure in browser
- Design primers for qPCR (avoid introns!)
- Check tissue expression
- Find regulatory elements
**3. Evolutionary Studies**:
- Compare gene across species
- View conservation track
- Identify conserved non-coding elements
- Study gene gain/loss events
**4. Drug Development**:
- Target gene for drug
- View isoforms (alternative splicing)
- Check tissue-specific expression
- Design specific inhibitors
### Genome Build Versions (Human)
Understanding **genome builds** is critical!
**Major human genome builds**:
**HG18** (NCBI Build 36, 2006):
- First "complete" draft
- Many gaps and errors
- Now obsolete - don't use!
**HG19** (GRCh37, 2009):
- Major improvement
- Used for ~10 years
- Still used in many studies
- Many databases still in HG19
**HG38** (GRCh38, 2013-present):
- Current standard
- Better assembly
- Fewer gaps
- Better representation of genetic diversity
- Includes alternate loci (variants)
**What changes between builds?**
- **Sequence corrections**: Errors fixed
- **Gap filling**: Unknown regions sequenced
- **Coordinate changes**: Same gene, different position!
- **New genes added**: ~200-500 genes updated per build
- **Removed sequences**: Some were artifacts
**Critical warning**:Gene X in HG19: chr17:43,044,295-43,125,483
Gene X in HG38: chr17:43,009,127-43,090,315
↑
Different coordinates!
**Always specify genome build in publications!**
**Liftover tools**:
- Convert coordinates between builds
- UCSC LiftOver tool
- Not always perfect - some regions can't convert!
### Why Genome Builds Keep Updating
**~25% of gene annotations change in 2 years!**
**Reasons for updates**:
**1. New experimental data**:
- More RNA-seq experiments
- New transcripts discovered
- Alternative splicing patterns refined
**2. Better computational methods**:
- Improved gene prediction algorithms
- Machine learning approaches
- Better integration of evidence
**3. Error corrections**:
- Previous annotations were wrong
- Pseudogenes misclassified as genes
- Gene boundaries corrected
**4. New regulatory elements**:
- ENCODE project found millions of regulatory regions
- Enhancers identified
- Non-coding RNAs discovered
**5. Improved diversity representation**:
- Original genome from few individuals
- Adding more diverse sequences
- Representing human variation better
**Future**: Pangenome (multiple reference genomes representing diversity)
## Databases: The Libraries of Biology
### Major Biological Databases
**GenBank (NCBI)**:
- DNA and RNA sequences
- Over 1 trillion bases!
- Publicly accessible
- Updated daily
**UniProt**:
- Protein sequences and functions
- Millions of proteins
- Annotated information
**PDB (Protein Data Bank)**:
- 3D protein structures
- X-ray crystallography and Cryo-EM data
- Visualize proteins
**Ensembl**:
- Genome browsers
- Gene annotations
- Comparative genomics
**KEGG (Kyoto Encyclopedia of Genes and Genomes)**:
- Metabolic pathways
- Gene functions
- Disease information
**Think of databases as**:
- Google for genes
- Wikipedia for proteins
- Library of Congress for genomes
## Sequence Alignment Algorithms
### Finding Similarities
**Why align sequences?**
- Find related genes
- Predict function
- Understand evolution
- Identify mutations
### BLAST: The Google of Bioinformatics
**BLAST** = Basic Local Alignment Search Tool [@altschul1990basic; @altschul1997gapped]
**What it does**:
- Takes your sequence
- Searches entire database
- Finds similar sequences
- Returns matches ranked by similarity
**Figure 17.2**: BLAST (Basic Local Alignment Search Tool) algorithm showing the seed-and-extend strategy for finding similar sequences in databases.
*Image credit: Wikimedia Commons, Public Domain*
**How fast?**
- Searches millions of sequences in seconds!
- Incredibly useful!
- Used millions of times per day worldwide!
### Types of BLAST
**Different BLAST programs for different tasks**:
| BLAST Type | Query | Database | Use Case |
|------------|-------|----------|----------|
| **BLASTN** | DNA/RNA | DNA/RNA | Find similar DNA sequences |
| **BLASTP** | Protein | Protein | Find similar proteins |
| **BLASTX** | DNA | Protein | Translate DNA, search proteins |
| **TBLASTN** | Protein | DNA | Search translated DNA with protein |
| **TBLASTX** | DNA | DNA | Both translated to protein first |
**When to use each**:
**BLASTN**:
- Identify species from DNA barcode
- Find orthologs in closely related species
- Map primers to genome
- Find contamination in sequencing
**BLASTP**:
- Predict protein function
- Find protein families
- Identify domains
- Evolutionary studies
**BLASTX**:
- Translate DNA in all 6 reading frames
- Search for protein homologs
- When you have DNA but want to find protein function
- Useful for ESTs, RNA-seq data
**Example use**:
- Found unknown gene in organism
- BLAST against database
- Find it matches insulin in mice
- Probably insulin in your organism too!
### Understanding BLAST Results
**Key metrics in BLAST output**:
**1. E-value (Expectation value)**:
- **Most important metric!**
- Probability of finding match by chance
- Lower = better!
**Interpretation**:
- **E < 1e-50**: Extremely significant match (almost certainly homologous)
- **E < 1e-10**: Very significant (likely homologous)
- **E < 0.01**: Significant (probably homologous)
- **E > 0.01**: Not significant (might be random)
**Example**:E-value = 1e-100: Expect 1 match by chance in 10^100 searches (excellent!)
E-value = 0.05: Expect 1 match by chance in 20 searches (weak)
**2. Bit Score**:
- Normalized alignment score
- Higher = better
- Independent of database size (unlike E-value)
- Use for comparing across databases
**3. Percent Identity**:
- Percentage of identical residues
- **Critical for functional inference!**
**Identity thresholds for function prediction**:
- **>70%**: Very likely same function
- **40-70%**: Possibly related function (use caution!)
- **<40%**: Uncertain relationship (different function likely)
**Example**:Query: Unknown protein from bacteria
Top BLAST hit:
- Description: DNA helicase
- E-value: 1e-85
- Identity: 78%
- Bit score: 320
Interpretation: Highly significant match (E-value), high identity (78%)
→ Query protein is likely a DNA helicase with similar function!
**4. Query Coverage**:
- Percentage of query sequence aligned
- Important for full-length matches
- Low coverage might indicate partial match or domain
**Best match characteristics**:
- Low E-value (< 1e-10)
- High identity (>70% for function)
- High coverage (>80%)
- High bit score
### Sequence Similarity and Functional Inference
**The similarity-function relationship**:
**Protein sequences**:
| % Identity | Functional Relationship | Confidence |
|------------|------------------------|------------|
| **>90%** | Same function, likely orthologs | **Very high** |
| **70-90%** | Same function, possible paralogs | **High** |
| **40-70%** | Related function, same family | **Moderate** |
| **25-40%** | Distant homologs, possibly different function | **Low** |
| **<25%** | Twilight zone - uncertain | **Very low** |
**DNA sequences** (more stringent):
| % Identity | Interpretation |
|------------|----------------|
| **>95%** | Same species, recent divergence |
| **85-95%** | Closely related species |
| **70-85%** | Related species, conserved genes |
| **<70%** | Distant relationship or chance |
**Important caveats**:
**1. Domain matches**:
- Proteins may share domains but have different functions
- Check if match is full-length or just domain!
**Example**:Query: 500 aa protein
BLAST match: Kinase domain (100 aa) at 80% identity
→ Query has kinase domain but overall function unclear
**2. Multidomain proteins**:
- Proteins with multiple domains
- Each domain may match different proteins
- Function depends on domain combination
**3. Paralog vs. Ortholog**:
- **Orthologs**: Same gene in different species (usually same function)
- **Paralogs**: Related genes from duplication (often different function)
- High similarity doesn't guarantee same function if paralogs!
### Gene Families and Homology
**Homology** = Similarity due to common evolutionary origin
**Types**:
**1. Orthologs**:
- Same gene in different species
- Arose from speciation
- **Usually same function**
- Best for functional transfer!
**Example**:Human β-globin ←→ Mouse β-globin (orthologs)
Both carry oxygen in red blood cells
**2. Paralogs**:
- Related genes within same genome
- Arose from gene duplication
- **Often diverged functions**
- Be careful transferring function!
**Example**:Human α-globin ←→ Human β-globin (paralogs)
Both in hemoglobin but slightly different roles
**3. Gene Families**:
- Groups of related genes
- Share common ancestor
- Varying degrees of functional similarity
**Examples of gene families**:
- **Globin family**: α-globin, β-globin, myoglobin
- **Immunoglobulin family**: Antibodies, T-cell receptors
- **HOX genes**: Developmental regulators
- **Kinase family**: Thousands of kinases with related activity
**Phylogenetic analysis** helps distinguish orthologs from paralogs:
1. BLAST to find homologs
2. Build phylogenetic tree
3. Identify ortholog groups
4. Transfer function to orthologs (not paralogs!)
### Homology-Based Functional Annotation
**Workflow for annotating unknown gene**:
**Step 1: BLAST Search**Query: Unknown gene from new organism
Database: NCBI nr (non-redundant protein)
Result: Top 100 hits
**Step 2: Examine Top Hits**Hit 1: DNA polymerase III (E=1e-120, ID=75%)
Hit 2: DNA polymerase III (E=1e-118, ID=74%)
Hit 3: DNA polymerase III (E=1e-115, ID=73%)
...
→ Consistent annotation!
**Step 3: Check Evidence Quality**
- Are top hits experimentally validated?
- Or just computational predictions?
- Look for "RefSeq", "SwissProt" (curated)
- Avoid "TrEMBL", "Predicted" (less reliable)
**Step 4: Examine Domains**
- Use InterPro, Pfam to find domains
- DNA polymerase has polymerase domain
- Confirms function!
**Step 5: Check GO Terms**
- DNA polymerase III:
- MF: DNA-directed DNA polymerase activity
- BP: DNA replication
- CC: Replisome
**Step 6: Assign Function with Confidence Level**Gene annotation: DNA polymerase III subunit alpha
Evidence: BLAST (E=1e-120, ID=75%), Pfam domain, GO terms
Confidence: HIGH
### Common BLAST Pitfalls
**Problem 1: Low complexity regions**
**Issue**: Repetitive sequences (e.g., poly-A, poly-Q) give false matches
**Solution**: Use low-complexity filter (default in BLAST)BLAST parameter: -seg yes (for proteins)
-dust yes (for DNA)
**Problem 2: Short queries**
**Issue**: Short sequences give many spurious hits
**Solution**:
- Need >30 aa for proteins, >50 bp for DNA
- Lower E-value threshold
- Use higher word size
**Problem 3: Database choice**
**Issue**: Wrong database gives misleading results
**Databases**:
- **nr**: All sequences (comprehensive but noisy)
- **RefSeq**: Curated, high quality (recommended!)
- **SwissProt**: Manually curated proteins (gold standard, but incomplete)
- **PDB**: Only sequences with 3D structures
- **Organism-specific**: Limit to taxonomic group
**Problem 4: Outdated databases**
**Issue**: Databases update frequently
**Solution**: Always use current database version!
**Problem 5: Misinterpreting paralogs as orthologs**
**Issue**: Transfer wrong function from paralog
**Solution**: Build phylogenetic tree to confirm orthology
### Advanced BLAST: PSI-BLAST
**PSI-BLAST** = Position-Specific Iterated BLAST
**What it does**:
- Iterative search
- Builds profile from first results
- Uses profile to find more distant homologs
- **Detects remote homologs** that regular BLAST misses!
**Workflow**:
1. Run initial BLAST
2. Build position-specific scoring matrix (PSSM) from hits
3. Search again with PSSM
4. Iterate 3-5 rounds
5. Finds distant family members!
**Use cases**:
- Find distant evolutionary relationships
- Expand protein families
- Detect remote homologs (<30% identity)
**Caution**: Can diverge and pick up false positives in later iterations!
### BLAST for Primer Design
**Use BLAST to check primer specificity**:
**Goal**: Ensure primers bind only to target gene
**Workflow**:
1. Design primers for gene of interest
2. BLAST each primer against genome
3. Check for off-target matches
4. Redesign if necessary
**Settings for primer BLAST**:
- Short query (18-25 bp primers)
- Allow some mismatches
- Check both forward and reverse primers
- Use organism-specific database
**Tool**: NCBI Primer-BLAST (specialized for this!)
### Gene Family Example: Actin
**Using BLAST to study gene families**:
**Query**: Human β-actin
**BLAST results** (simplified):
1. Mouse β-actin (E=0, ID=99%) ← Ortholog
2. Rat β-actin (E=0, ID=98%) ← Ortholog
3. Human α-actin (E=1e-180, ID=93%) ← Paralog
4. Yeast actin (E=1e-150, ID=88%) ← Ancient ortholog
5. Plant actin (E=1e-140, ID=85%) ← Ancient ortholog
**Interpretation**:
- β-actin highly conserved across mammals (orthologs)
- α-actin and β-actin are paralogs (gene duplication)
- Actin family ancient (found in yeast, plants!)
- All have similar function (cytoskeleton)
### BLAST Databases
**Major databases**:
**1. NCBI nr (non-redundant)**:
- All GenBank + RefSeq + PDB + SwissProt
- Most comprehensive
- Updated daily
- Some redundancy despite name!
**2. RefSeq**:
- Curated reference sequences
- High quality annotations
- Non-redundant
- **Recommended for most uses**
**3. SwissProt** (UniProtKB/Swiss-Prot):
- Manually curated
- Highest quality
- Experimental evidence
- But incomplete coverage
**4. TrEMBL** (UniProtKB/TrEMBL):
- Computationally annotated
- Comprehensive but lower quality
- Supplement to SwissProt
**5. PDB**:
- Only proteins with 3D structures
- Small but high-quality
- Use for structural comparisons
**6. Organism-specific**:
- Human, mouse, E. coli, etc.
- Faster searches
- More relevant results
### Interpreting BLAST for Pathogen Identification
**Use case**: Unknown bacteria, want to identify species
**Workflow**:
1. Sequence 16S rRNA gene (universal barcode)
2. BLAST against 16S rRNA database
3. Top hit identifies species
**Example**:Query: 16S rRNA from unknown bacteria
Top BLAST hit:
- Escherichia coli str. K-12 (E=0, ID=99.8%)
Interpretation: Sample is E. coli K-12 strain
**Critical thresholds** (16S rRNA):
- **>99%**: Same species
- **97-99%**: Likely same species
- **95-97%**: Same genus
- **<95%**: Different genus
### BLAST Best Practices
**Do's**:
✅ Use appropriate BLAST type (BLASTN vs. BLASTP)
✅ Check E-value AND percent identity
✅ Examine multiple top hits (consensus annotation)
✅ Use curated databases when possible (RefSeq, SwissProt)
✅ Consider query coverage
✅ Distinguish orthologs from paralogs
✅ Check domain architecture
✅ Keep databases updated
**Don'ts**:
❌ Don't rely on single BLAST hit
❌ Don't ignore E-value
❌ Don't assume high similarity = same function (check for paralogs!)
❌ Don't use outdated databases
❌ Don't transfer function from distant homologs (<40% identity)
❌ Don't forget to filter low-complexity regions
### Pairwise Alignment
**Comparing two sequences**:
**Global alignment** (Needleman-Wunsch):
- Align entire sequences end-to-end
- Best for similar-length, similar sequences
**Local alignment** (Smith-Waterman):
- Find best matching regions
- Don't need to align everything
- BLAST uses this approach!
### Multiple Sequence Alignment
**Aligning many sequences at once**:
- Finds conserved regions across species
- Reveals important functional sites
- Used for phylogenetics
**Tools**: MUSCLE, MAFFT, Clustal Omega
## Sequence Conservation and Evolutionary Analysis
### What Is Sequence Conservation?
**Conservation** = How similar a DNA/protein sequence is across different species or within a genome [@siepel2005evolutionarily; @pollard2010detection].
Think of it like:
- 📖 **Ancient texts** preserved through copying (important texts conserved more carefully)
- 🏛️ **Historical buildings** maintained over centuries (functional ones preserved)
- 💎 **Family heirlooms** passed down unchanged (valuable things kept safe)
**The principle**: **If a sequence is conserved, it's probably important!**
**Why?**
- Natural selection eliminates harmful mutations
- Functional sequences resist change
- Non-functional sequences accumulate random mutations
- Evolution is the ultimate experiment!
### Why Conservation Matters
**Conservation reveals**:
**1. Functional Importance**:
- Highly conserved → functionally critical
- Variable regions → less critical
- Like finding which parts of a machine are essential
**2. Regulatory Elements**:
- Conserved non-coding sequences often regulatory
- Enhancers, promoters, silencers
- Hard to find by sequence alone
- Conservation highlights them!
**3. Disease-Causing Mutations**:
- Mutations in conserved regions more likely pathogenic
- Used in clinical variant interpretation
- Conservation score predicts impact
**4. Drug Targets**:
- Target conserved regions in pathogens
- Avoid conserved regions in humans
- Reduces side effects
### Measuring Conservation
**Conservation score** = Quantitative measure of sequence similarity
**Range**: Typically 0 to 1 (or 0 to 100%)
- **1.0** (100%) = Perfectly conserved (identical across species)
- **0.5** (50%) = Moderately conserved
- **0.0** (0%) = Not conserved (random variation)
**Popular methods**:
**1. PhyloP**:
- Tests for conservation vs. neutral evolution
- Positive scores = conserved
- Negative scores = accelerated evolution
- Based on phylogenetic trees
**2. PhastCons**:
- Identifies conserved elements
- Hidden Markov Model approach
- Returns probability of conservation
- Smooths across neighboring bases
**3. GERP (Genomic Evolutionary Rate Profiling)**:
- Detects constrained elements
- Compares observed vs. expected substitutions
- Positive scores = conserved
- Widely used in clinical genomics
### Conservation Across Genome Regions
Different parts of genes show different conservation levels:
| Region | Conservation | Why? |
|--------|--------------|------|
| **Exons** (coding) | **Very high** | Changes alter protein sequence |
| **Start/Stop codons** | **Extremely high** | Essential for translation |
| **Promoters** | **High** | Required for transcription |
| **Splice sites** | **Very high** | GT...AG boundaries critical |
| **Enhancers** | **Moderate to high** | Functional but flexible |
| **Introns** | **Low to moderate** | Some regulatory elements |
| **5' and 3' UTRs** | **Moderate** | Regulatory sequences present |
| **Intergenic** | **Low** | Mostly non-functional |
**Example - BRCA1 gene**:Conservation track (UCSC Genome Browser):
Exon 11: ████████████ (0.95 - highly conserved)
Intron 10: ▒▒▒░░░▒▒░░░ (0.35 - less conserved)
Promoter: ████▒▒████ (0.80 - conserved)
**Interpretation**: Exons and promoter are conserved (functional), introns less so
### Conservation and Codon Usage
**Synonymous vs. Non-synonymous mutations**:
**Synonymous (silent) mutations**:
- Change DNA codon but **NOT** amino acid
- Example: CTT → CTC (both code for Leucine)
- Often tolerated (lower selection pressure)
- Lower impact on conservation
**Non-synonymous mutations**:
- Change DNA codon AND amino acid
- Example: CTT (Leu) → CAT (His)
- Often deleterious if in conserved region
- Higher selection pressure against them
**Conservation scoring accounts for this**:
- Non-synonymous changes in conserved regions → high pathogenicity score
- Synonymous changes → lower pathogenicity score
- Transition mutations (purine ↔ purine, pyrimidine ↔ pyrimidine) more tolerated than transversions
### Example: Actin Conservation
**Actin** is one of the most conserved proteins!
**Comparison**:
- Yeast actin vs. Human actin
- **Sequence similarity: ~90%!**
- Separated by >1 billion years of evolution
- Conservation score near 1.0 across entire gene
**Why so conserved?**
- Essential for cell structure
- Interacts with many proteins
- Changes break cellular machinery
- Strong negative selection
**Practical implication**: Can study human actin in yeast! (Model organisms)
### Example: UTR Conservation
**5' and 3' UTRs** (Untranslated Regions):
**Generally less conserved than exons**:
- Exons: 0.85-0.95 conservation
- UTRs: 0.50-0.70 conservation
**But important exceptions**:
- microRNA binding sites in 3' UTR → highly conserved
- Regulatory elements in 5' UTR → conserved
- Shows functional importance!
**Practical use**:
1. Sequence gene across species
2. Align UTRs
3. Find conserved patches
4. → Likely regulatory elements!
### PhyloP and PhastCons Scores
**PhyloP** (Phylogenetic P-values):
**What it measures**:
- Conservation or acceleration at each position
- Based on multiple species alignment
- Statistical significance of conservation
**Score interpretation**:
- **PhyloP > +2.0**: Highly conserved (p < 0.05)
- **PhyloP = 0**: Neutral evolution
- **PhyloP < -2.0**: Accelerated evolution (positive selection?)
**PhastCons** (Phylogenetic Hidden Markov Model Conservation):
**What it measures**:
- Probability that each position is in a conserved element
- Considers neighboring positions (smoothing)
- Better for finding conserved regions (not just bases)
**Score interpretation**:
- **PhastCons > 0.8**: Likely in conserved element
- **PhastCons 0.3-0.8**: Moderate conservation
- **PhastCons < 0.3**: Not conserved
**Visualizing in genome browser**:UCSC Genome Browser → Add Track → Conservation
- Vertebrate Conservation (100 species)
- Mammal Conservation (60 species)
- Primate Conservation (20 species)
View PhyloP and PhastCons tracks side-by-side!
### Using Conservation in Variant Interpretation
**Clinical genetics workflow**:
**Step 1**: Patient has variant in gene
**Step 2**: Check conservation at that position
**Step 3**: Interpret:
| Conservation Score | Interpretation | Likelihood of Pathogenicity |
|--------------------|----------------|----------------------------|
| **PhyloP > 5** | Extremely conserved | **Very high** - likely pathogenic |
| **PhyloP 2-5** | Highly conserved | **High** - likely damaging |
| **PhyloP 0-2** | Moderately conserved | **Moderate** - uncertain |
| **PhyloP < 0** | Not conserved | **Low** - likely benign |
**Example**:Patient variant: BRCA1 c.5266dupC (frameshift)
Position: Exon 20
PhyloP score: 6.8 (extremely conserved)
Interpretation: Highly conserved region → variant likely pathogenic
Clinical action: High-risk cancer surveillance
**ACMG Guidelines** (Clinical variant interpretation):
- Conservation is **"Supporting" evidence** for pathogenicity
- Combined with other evidence (functional studies, population frequency, etc.)
- Not used alone!
### Conserved Non-Coding Elements
**Surprising discovery**: Some non-coding regions extremely conserved!
**Examples**:
**1. Ultraconserved Elements (UCEs)**:
- **100% identical** across human, mouse, rat
- Often >200 bp long
- Function often unknown!
- Likely critical regulatory elements
**2. Conserved Non-coding Sequences (CNS)**:
- High conservation in intergenic regions
- Enhancers for developmental genes
- Many near Hox genes, transcription factors
**3. MicroRNA binding sites**:
- Short conserved sequences in 3' UTRs
- ~7 nucleotides (seed region)
- Critical for gene regulation
**Discovery method**:
1. Align human genome to 100 vertebrate genomes
2. Find conserved regions outside genes
3. Test in experiments (reporter assays)
4. → Many are enhancers!
**Clinical importance**: Mutations in these regions can cause disease!
**Example**: Mutations in sonic hedgehog (SHH) enhancer cause limb malformations, even though SHH gene itself is normal!
### Evolutionary Rate and Selection
**Rate of evolution** varies by selective pressure:
**Strong purifying selection** (negative selection):
- Removes harmful mutations
- **Result**: High conservation
- **Example**: Active site of enzymes
**Neutral evolution**:
- No selection pressure
- **Result**: Low conservation
- **Example**: Synonymous sites, intergenic regions
**Positive selection**:
- Favors new mutations
- **Result**: Accelerated evolution (low conservation, but meaningful!)
- **Example**: Immune system genes (fighting ever-changing pathogens)
**dN/dS ratio** (Non-synonymous/Synonymous substitution rate):
- **dN/dS < 1**: Purifying selection (conserved)
- **dN/dS = 1**: Neutral evolution
- **dN/dS > 1**: Positive selection (adaptive)
### Conservation-Based Gene Finding
**Strategy**: Use conservation to find genes!
**Approach**:
1. Align genome to related species
2. Find conserved ORFs
3. → Likely protein-coding genes!
**Advantages**:
- Confirms ab initio predictions
- Finds genes missed by other methods
- Reduces false positives
**Example**:Ab initio prediction: 100 potential genes
Conservation filtering: 82 conserved ORFs
→ High confidence: 82 genes
→ Low confidence: 18 genes (likely false positives)
### Mutation Types and Conservation Impact
**Impact varies by mutation type and conservation**:
| Mutation Type | Conserved Region | Non-Conserved Region |
|---------------|------------------|----------------------|
| **Synonymous** | Low impact | Very low impact |
| **Missense** (amino acid change) | **High impact** | Low-moderate impact |
| **Nonsense** (stop codon) | **Very high impact** | Moderate impact |
| **Frameshift** | **Extremely high impact** | Moderate-high impact |
| **Splice site** | **Extremely high impact** | Moderate impact (if in conserved intron) |
**Practical application**: Prioritize variants in conserved regions for follow-up studies
### Tools for Conservation Analysis
**Genome browsers**:
- **UCSC Genome Browser**: PhyloP, PhastCons tracks
- **Ensembl**: Conservation scores
- **IGV**: Load conservation tracks
**Variant effect predictors**:
- **CADD** (Combined Annotation Dependent Depletion): Integrates conservation
- **SIFT**: Uses sequence conservation to predict impact
- **PolyPhen-2**: Considers conservation in pathogenicity scoring
**Command-line tools**:
- **phyloP**: Calculate conservation scores
- **phastCons**: Identify conserved elements
- **GERP++**: Genomic conservation scoring
### Limitations of Conservation Analysis
**Important caveats**:
**1. Species-Specific Functions**:
- Not all functional elements are conserved
- Recent evolutionary innovations not conserved
- Human-specific regulatory elements missed
**2. Incomplete Sampling**:
- Conservation depends on which species compared
- More species → better power
- But bias toward well-sequenced organisms
**3. Neutral Conserved Regions**:
- Low mutation rate ≠ functional importance
- Some regions conserved by chance
- Need experimental validation
**4. Rapidly Evolving Functional Elements**:
- Some genes under positive selection
- Immune genes, olfactory receptors
- Low conservation doesn't mean non-functional!
### Best Practices
**Do's**:
✅ Use conservation as **one line of evidence**, not sole criterion
✅ Consider multiple species alignments
✅ Use appropriate conservation metric (PhyloP, PhastCons, GERP)
✅ Account for mutation type (synonymous vs. non-synonymous)
✅ Validate with experimental data when possible
**Don'ts**:
❌ Don't assume low conservation = non-functional
❌ Don't ignore species-specific elements
❌ Don't use conservation alone for clinical decisions
❌ Don't forget that conservation evolves (some elements lost/gained)
## Practical Application: Mutation and Disease Analysis
### Integrating Bioinformatics for Clinical Genomics
**Workflow**: Patient with suspected genetic disease
**Goal**: Identify disease-causing mutation and understand mechanism
### Step-by-Step Clinical Analysis
**Step 1: Variant Detection from Sequencing**Patient DNA → Whole exome/genome sequencing → Variant calling
Result: List of ~20,000-100,000 variants (SNPs, indels) per patient
**Challenge**: Which variant causes disease?
**Step 2: Variant Filtering**
**Apply computational filters**:
1. **Frequency filter**: Remove common variants (MAF > 1%)
- Disease-causing variants usually rare
- Use gnomAD database
- Reduces to ~1,000 variants
2. **Gene filter**: Focus on disease-relevant genes
- Known disease genes
- Genes in relevant pathway
- Reduces to ~100-500 variants
3. **Functional impact filter**: Prioritize high-impact variants
- Nonsense, frameshift → HIGH
- Missense in conserved region → MODERATE
- Synonymous → LOW
- Reduces to ~10-50 candidates
**Step 3: Variant Annotation**
**For each candidate variant, determine**:
**A. Gene and Region Affected**
- Which gene?
- Exon, intron, UTR, promoter?
- Use genome browser (UCSC, Ensembl)
**B. Functional Prediction**
- BLAST homology: Is position conserved?
- PhyloP conservation score
- SIFT/PolyPhen prediction (pathogenic vs. benign)
**C. Existing Knowledge**
- Check ClinVar database
- Has this variant been reported?
- Known pathogenic/benign classification?
**D. Gene Ontology**
- What pathways affected?
- Compatible with patient phenotype?
**Step 4: Detailed Analysis of Top Candidate**
**Example case**:Patient: 8-year-old with cardiomyopathy (heart muscle disease)
Sequencing: Exome sequencing
Candidate variant: MYH7 c.2389C>T (p.Arg797Cys)
**Annotation workflow**:
**1. Gene identification**:
- Gene: MYH7 (myosin heavy chain 7)
- Location: Chromosome 14
- Function: Cardiac muscle contraction
**2. ORF and protein impact**:
- ORF disruption: Missense mutation (Arg → Cys)
- Position: Codon 797 in exon 21
- Protein domain: Motor domain (critical!)
**3. Conservation analysis**:
- BLAST MYH7 across species:
- Human-Mouse: 98% identity
- Human-Zebrafish: 85% identity
- Position 797 (Arg):
- PhyloP score: 7.2 (highly conserved!)
- 100% conserved across all vertebrates
- → Mutation at highly conserved position
**4. Functional prediction**:
- SIFT score: 0.01 (deleterious, cutoff 0.05)
- PolyPhen-2: 0.98 (probably damaging, cutoff 0.85)
- CADD score: 28 (pathogenic, cutoff 20)
- → All predictors agree: pathogenic!
**5. Gene Ontology**:
- MF: Motor activity (GO:0003774)
- BP: Cardiac muscle contraction (GO:0060048)
- CC: Myosin complex (GO:0016459)
- → Gene function matches patient phenotype!
**6. ClinVar check**:
- Variant reported 15 times
- Classification: **Pathogenic**
- Associated disease: Hypertrophic cardiomyopathy
- → Known disease-causing variant!
**7. Literature review**:
- PubMed search: MYH7 + cardiomyopathy
- Multiple papers confirm pathogenicity
- Functional studies show disrupted muscle contraction
**Conclusion**: MYH7 p.Arg797Cys is disease-causing variant
**Clinical action**: Genetic counseling, family screening, treatment plan
### Mutation Analysis: Promoter Regions
**Scenario**: Mutation in promoter region
**Challenge**: Promoters not translated → no protein change, but affects expression
**Analysis approach**:
**1. Identify promoter elements**:
- Use genome browser (UCSC)
- Look for CpG islands, TATA box, transcription factor binding sites
- Check conservation across species
**2. Assess mutation location**:
- Is mutation in core promoter? (critical!)
- In enhancer region?
- In transcription factor binding site?
**3. Predict transcriptional impact**:
- Mutations in TATA box → reduced transcription
- Mutations in enhancer → tissue-specific effects
- Use tools: JASPAR (TF binding site prediction)
**4. Experimental validation**:
- Reporter assays (luciferase)
- Clone promoter with/without mutation
- Measure expression level
- Validate computational predictions!
**Example**:Variant: β-globin promoter mutation (-28 A>G)
Effect: Reduced transcription by 70%
Disease: β-thalassemia (low hemoglobin)
Mechanism: Disrupts TATA box recognition
### Mutation Analysis: UTR Regions
**5' UTR mutations**:
- Affect translation efficiency
- Can create upstream ORFs (uORFs)
- Interfere with ribosome scanning
**3' UTR mutations**:
- Affect mRNA stability
- Disrupt microRNA binding sites
- Change polyadenylation
**Analysis**:
1. Check conservation (conserved UTR patches → regulatory!)
2. Predict microRNA binding sites (TargetScan, miRanda)
3. Assess polyadenylation signals
4. Reporter assays to measure mRNA stability
### Drug Target Discovery Workflow
**Goal**: Find new antibiotic targets against pathogenic bacteria
**Strategy**: Target essential bacterial genes not found in humans
**Workflow**:
**Step 1: Sequence pathogen genome**Example: New drug-resistant Staphylococcus aureus strain
→ Whole genome sequencing
→ Gene annotation (ab initio + homology)
**Step 2: Identify essential genes**
- Literature: Known essential genes in related bacteria
- Experiments: Transposon mutagenesis screens
- Result: ~300 essential genes
**Step 3: Filter for bacterial-specific genes**For each essential gene:
BLAST against human genome
If high similarity (>50%) → EXCLUDE (would affect human!)
If no/low similarity → KEEP as candidate
Result: ~150 bacterial-specific essential genes
**Step 4: GO term analysis**
- Group by function:
- Cell wall synthesis (30 genes)
- DNA replication (25 genes)
- Protein synthesis (40 genes)
- Metabolism (55 genes)
**Step 5: Prioritize druggable targets**
- Enzymes > structural proteins (easier to inhibit)
- Look for active sites, binding pockets
- Check if similar targets already drugged
- Result: ~20 high-priority targets
**Step 6: Conservation across bacterial strains**BLAST each target across:
- Multiple S. aureus strains
- Related bacteria (broad-spectrum potential)
Highly conserved (>90%) → Broad-spectrum antibiotic!
Variable → Strain-specific target
**Step 7: Experimental validation**
- Synthesize/purify target protein
- Screen chemical libraries for inhibitors
- Test in bacterial growth assays
- Lead optimization
**Example outcome**:Target: FabI (enoyl-ACP reductase)
Function: Fatty acid synthesis (essential)
GO terms: Lipid metabolic process
Conservation: 95% across Staph species
Human homolog: Different enzyme (safe!)
Drug: Triclosan (proof of concept)
Status: Validated antibiotic target!
### Annotation Best Practices Summary
From the lecture notes, **key principles for robust annotation**:
**1. Cross-validation**:
- Use multiple algorithms (ab initio + homology + RNA-seq)
- Compare results
- Trust consensus predictions
**2. Avoid automation errors**:
- Computational predictions not perfect
- Manual curation essential for important genes
- Expert review reduces errors
**3. Account for alternative splicing**:
- Multiple isoforms from one gene
- Different proteins, different functions
- Annotate all isoforms
**4. Document methods clearly**:
- Which tools used?
- Which databases?
- Which genome build (HG19 vs HG38)?
- Essential for reproducibility!
**5. Use multiple evidence types**:
- Sequence similarity (BLAST)
- Conservation (PhyloP/PhastCons)
- Expression data (RNA-seq)
- Protein evidence (mass spec)
- Literature (PubMed)
**6. Uncertainty handling**:
- If uncertain → mark as "hypothetical protein"
- Better than wrong annotation!
- Provides confidence scores when possible
### Integration Example: Full Workflow
**Scenario**: Novel gene in newly sequenced organism
**Complete annotation pipeline**:
**1. ORF Detection**:
- Scan for start codon (ATG)
- Find stop codons
- Identify longest ORF
- Check for introns (if eukaryote)
**2. Homology Search (BLAST)**:
- BLASTP against RefSeq
- Top hit: 75% identity to DNA helicase
- E-value: 1e-120 (highly significant)
**3. Domain Analysis**:
- InterPro/Pfam scan
- Detects: Helicase domain (IPR001650)
- Confirms BLAST prediction
**4. Conservation**:
- PhyloP: 6.5 (highly conserved)
- Aligns across 50 species
- Critical residues 100% conserved
**5. Gene Ontology**:
- MF: DNA helicase activity (GO:0003678)
- BP: DNA replication (GO:0006260)
- CC: Nucleus (GO:0005634)
**6. Expression (if available)**:
- RNA-seq: Highly expressed in dividing cells
- Consistent with replication function
**7. Final Annotation**:Gene: helicase-1 (hel1)
Product: DNA helicase, putative
Function: DNA replication
Evidence: BLAST (E=1e-120, ID=75%), Pfam domain,
conservation (PhyloP=6.5), expression pattern
Confidence: HIGH
GO: 0003678, 0006260, 0005634
**Quality control**:
- Multiple lines of evidence agree ✓
- High confidence annotation ✓
- Well-documented ✓
- Ready for publication/database submission ✓
## Primer Design and Quantitative PCR (qPCR)
### What Are Primers?
**Primers** = Short DNA sequences (18-25 nucleotides) that bind to template DNA and initiate DNA synthesis
Think of primers like:
- 🎯 **GPS coordinates** telling DNA polymerase where to start
- 🔑 **Keys** that unlock specific DNA regions for amplification
- 🏁 **Starting flags** marking where copying begins
**Why we need them**:
- DNA polymerase cannot start synthesis de novo (from scratch)
- Needs 3'-OH group from existing nucleotide
- Primers provide this starting point!
**Historical note**: This is why DNA polymerase differs from RNA polymerase (which CAN start de novo)
### Primer Design Principles
**Critical parameters** for successful primers:
**1. Length**:
- **Optimal**: 18-25 nucleotides
- **Too short** (<15 bp): Not specific enough (binds multiple sites)
- **Too long** (>30 bp): Expensive, slow annealing
**2. Melting Temperature (Tm)**:
- Temperature at which 50% of primers bind to template
- **Optimal**: 55-65°C
- **Forward and reverse primers should have similar Tm** (within 2-3°C)
- Calculate using formula or online tools
**Basic Tm calculation**:Tm = 4(G+C) + 2(A+T) (rough estimate)
Example primer: 5'-ATGCGTACGGATCCGTAA-3'
A=4, T=3, G=5, C=6
Tm = 4(5+6) + 2(4+3) = 44 + 14 = 58°C
**More accurate**: Use Nearest-Neighbor method or online calculators
**3. GC Content**:
- Percentage of G and C nucleotides
- **Optimal**: 40-60%
- **Too low** (<30%): Weak binding
- **Too high** (>70%): Too strong binding, non-specific
**4. Specificity**:
- Primer should bind ONLY to target sequence
- Check with BLAST (Primer-BLAST tool)
- Avoid repetitive sequences
- Avoid homopolymeric runs (AAAA, GGGG, etc.)
**5. Avoiding Secondary Structures**:
**Self-complementarity**:
- Primer shouldn't bind to itself
- Causes primer-dimers (waste primers!)
Bad: 5'-ATGCGCGCAT-3'
||||||||| (palindrome - binds to itself!)
**Hairpins**:
- Intramolecular folding
- Prevents primer from binding template
Bad primer with hairpin:
5'-ATGCGGCCGCAT-3'
↓ ↑
====== (stem-loop structure)
**6. 3' End Stability**:
- Last 5 nucleotides at 3' end critical
- Should have 1-2 G or C (but not all GCs!)
- **Avoid**: Multiple Gs/Cs at 3' end (binds non-specifically)
### Primer Design for Different Applications
### A. PCR Amplification
**Goal**: Amplify specific DNA region
**Design strategy**:
1. Identify target region (gene, exon, etc.)
2. Design forward primer at 5' end of region
3. Design reverse primer at 3' end (complement!)
4. Product size: 100-1000 bp ideal
**Example**:Target: Amplify exon 5 of BRCA1 gene
5' ====[Exon 5]====3'
↑ ↑
FP RP
Forward primer (FP): Binds at start of exon
Reverse primer (RP): Binds at end (reverse complement!)
Product: Complete exon 5 sequence
**Using genome browser for primer design**:
1. Open UCSC Genome Browser
2. Navigate to target gene
3. Zoom to region of interest
4. View sequence (Tools → View DNA)
5. Extract sequences for primer design
6. **Avoid introns!** (primers should span exon-exon junctions or be in exons only)
### B. qPCR (Quantitative PCR / Real-Time PCR)
**Goal**: Measure gene expression levels (how much mRNA?)
**Different from regular PCR**:
- Measures product in **real-time** (not just endpoint)
- Quantitative (not just yes/no)
- Uses fluorescent dyes or probes
**Primer design for qPCR is MORE STRINGENT**:
**Key differences**:
**1. Product size**:
- **Optimal**: 80-150 bp (shorter than regular PCR!)
- Short amplicons = more efficient qPCR
**2. Avoid genomic DNA contamination**:
- Design primers across exon-exon junctions
- Or use DNase treatment
**Example**:Gene structure:
[Exon 1]--intron--[Exon 2]--intron--[Exon 3]
Bad design (amplifies genomic DNA too):
FP in Exon 1 → RP in Exon 1
(genomic DNA and cDNA both amplified)
Good design (cDNA specific):
FP in Exon 1 → RP in Exon 2
(spans intron - only cDNA amplified!)
Genomic DNA gives no product (intron too large)
cDNA gives short product (introns removed)
**3. No template controls essential**:
- Include wells with no template
- Check for primer-dimers
**4. Reference genes**:
- Need housekeeping gene for normalization
- GAPDH, β-actin, 18S rRNA
- Should have stable expression
### qPCR Workflow
**Step 1: RNA Extraction**
- Extract total RNA from cells/tissue
- Measure concentration (NanoDrop, Qubit)
**Step 2: Reverse Transcription (RT)**
- Convert RNA → cDNA using reverse transcriptase
- Now have DNA template for qPCR
**Step 3: qPCR Reaction**
- Mix cDNA + primers + fluorescent dye (e.g., SYBR Green)
- Run thermal cycles
- Monitor fluorescence in real-time
**Step 4: Data Analysis**
- Ct value (Cycle threshold): Cycle at which fluorescence exceeds background
- Lower Ct = more template (higher expression!)
- **ΔΔCt method** for quantification
**Ct interpretation**:Gene A in control cells: Ct = 20
Gene A in treated cells: Ct = 17
Difference = 3 cycles
Each cycle = 2x amplification
2^3 = 8-fold higher expression in treated cells!
### Primer Design Tools
**Online tools** (free and widely used):
**1. NCBI Primer-BLAST**:
- URL: https://www.ncbi.nlm.nih.gov/tools/primer-blast/
- Best for specificity checking
- BLAST primers against genome
- Finds potential off-target sites
**Workflow**:
1. Enter target sequence or gene name
2. Specify primer parameters (size, Tm, GC%)
3. Tool designs primers automatically
4. Checks specificity by BLAST
5. Returns primers with no off-targets!
**2. Primer3**:
- URL: https://primer3.ut.ee/
- Widely used, flexible
- Many parameters customizable
- Good for general PCR
**3. IDT PrimerQuest**:
- From Integrated DNA Technologies
- Commercial but free to use
- Good for qPCR primer design
**4. UCSC In-Silico PCR**:
- Test if primers amplify expected product
- Uses genome sequence
- Predicts product size
### Practical Example: Design qPCR Primers for BRCA1
**Goal**: Measure BRCA1 expression by qPCR
**Step 1: Open UCSC Genome Browser**Search: BRCA1
→ Chr 17: 43,000,000-43,100,000
View gene structure
**Step 2: Choose target region**
- Exons 10-11 (frequently expressed, large exons)
- Design primers spanning exon junction
**Step 3: Extract sequences**Exon 10: ...GCTATGAAGAATGGAAG...
Exon 11: ...AATGCCAAGAACTATGC...
Junction: GAATGGAAG|AATGCCAAG
Exon 10 | Exon 11
**Step 4: Design primers**Forward primer (spans junction):
5'-GAATGGAAGAATGCCAAG-3'
(last 5 bp of exon 10 + first 13 bp of exon 11)
Tm: 58°C, GC: 44%
Reverse primer (in exon 11):
5'-GCATAGTTCTTGGCATTC-3' (reverse complement)
Tm: 58°C, GC: 50%
Product size: 120 bp (ideal for qPCR!)
**Step 5: Check specificity with Primer-BLAST**
- No off-target amplification ✓
- Only hits BRCA1 ✓
- Ready to order!
**Step 6: Experimental validation**
- Test primers with positive control (BRCA1 cDNA)
- Check melt curve (single peak = specific!)
- Confirm product size by gel electrophoresis
### Troubleshooting Primer Problems
**Problem 1: No amplification**
**Causes**:
- Primers don't bind (wrong sequence)
- Template degraded
- Annealing temperature too high
**Solutions**:
- Check primer sequences
- Fresh template
- Gradient PCR (test range of temperatures)
**Problem 2: Multiple bands**
**Causes**:
- Non-specific primer binding
- Primer-dimers
**Solutions**:
- Redesign primers (higher specificity)
- Increase annealing temperature
- Touchdown PCR protocol
**Problem 3: Primer-dimers in qPCR**
**Cause**:
- Primers bind to each other instead of template
**Solution**:
- Redesign to avoid complementarity
- Reduce primer concentration
- Check no-template control
**Problem 4: Variable Ct values (qPCR)**
**Causes**:
- Pipetting errors
- RNA quality issues
- Inhibitors in reaction
**Solutions**:
- Careful pipetting (use repeat pipettors)
- Check RNA integrity (RIN score)
- Dilute template (reduce inhibitors)
### Expression vs. Amplification Primer Design
**Key difference**: Where primers bind matters for purpose!
**For expression quantification (qPCR)**:
- Primers must be **mRNA-specific**
- Span exon-exon junctions
- Avoid amplifying genomic DNA
- In coding region of gene
**For gene verification/amplification**:
- Primers can be anywhere in gene
- Introns OK if amplifying genomic DNA
- Promoter region OK for regulatory studies
**Example**:Quantify BRCA1 expression (qPCR):
→ Primers span exon 10-11 junction
→ Only amplifies cDNA (mRNA-derived)
Amplify BRCA1 promoter (regular PCR):
→ Primers in promoter region
→ Amplifies genomic DNA
→ Use for mutation screening```
21.3.2 Best Practices for Primer Design
Do’s: ✅ Use online tools (Primer3, Primer-BLAST) ✅ Check specificity with BLAST ✅ Design primers with similar Tm ✅ Avoid secondary structures ✅ For qPCR: span exon junctions ✅ Include positive and negative controls ✅ Validate experimentally
Don’ts: ❌ Don’t use primers with >3 G/C at 3’ end ❌ Don’t ignore secondary structure prediction ❌ Don’t skip BLAST specificity check ❌ Don’t design primers in repetitive regions ❌ Don’t use primers from different species (unless checking conservation)
21.4 Structural Bioinformatics
21.4.1 Predicting Protein Structures
Why it matters:
Structure determines function
Drug design needs structures
Understanding disease mutations
The problem:
Experiments are slow and expensive
Can we predict structure from sequence?
21.4.2 AlphaFold: AI Solves 50-Year Problem!
AlphaFold (Google DeepMind, 2020):
Uses deep learning / AI
Predicts 3D structure from sequence
Near-experimental accuracy!
Revolutionary!
Impact:
Predicted structures for 200+ million proteins!
Free database (AlphaFold DB)
Accelerating drug discovery
Nobel Prize potential
21.4.3 Homology Modeling
Approach:
If protein A’s structure is known
And protein B is similar sequence
Model B based on A’s structure
Works well for similar proteins!
21.4.4 Protein-Protein Docking
Predicting how proteins interact:
Important for understanding cell signaling
Drug development
Protein engineering
21.5 Machine Learning in Biology
21.5.1 AI Meets Genomics
Machine learning = Teaching computers to learn patterns from data
Applications:
1. Disease Prediction:
Analyze genomes to predict disease risk
Better than humans at finding subtle patterns
Personalized risk scores
2. Variant Interpretation:
Millions of variants in each genome
Which are disease-causing?
ML helps classify them
3. Drug Discovery:
Predict which molecules bind to targets
Design new drugs
Much faster than traditional methods
4. Image Analysis:
Analyze microscopy images
Count cells automatically
Detect cancer in pathology slides
5. Gene Regulation:
Predict which sequences control genes
Understand regulatory code
Design synthetic promoters
21.5.2 Deep Learning Revolution
Deep learning = Advanced ML using neural networks
Successes:
AlphaFold (protein structure)
DeepVariant (variant calling)
BaseNJ (DNA sequencing accuracy)
Drug response prediction
Future: AI will be essential partner in biology!
21.6 Genomic Data Formats
21.6.1 Standard File Formats
FASTA:
Sequence format
Simple text file
Header
ATCGATCG…
FASTQ:
Sequence + quality scores
Raw sequencing data
SAM/BAM:
Aligned sequences
Maps reads to genome
BAM is compressed SAM
VCF (Variant Call Format):
Lists genetic variants
Position, reference, alternate
Standard for sharing variants
GFF/GTF:
Gene annotations
Where genes are located
Exons, introns, etc.
21.6.2 Why Standards Matter
Benefits:
Different tools work together
Share data easily
Reproducible research
Collaborate globally
21.7 Programming in Bioinformatics
21.7.1 Common Languages
Python:
Most popular in bioinformatics
Easy to learn
Powerful libraries (BioPython)
Great for data analysis
R:
Statistical computing
Excellent for genomics
BioConductor (huge package collection)
Beautiful visualizations
Perl:
Text processing
Older but still used
BioPerl
Unix/Linux Command Line:
Essential skill!
File manipulation
Running tools
Automating workflows
Typical workflow:
Process data with Unix commands
Analyze with Python/R
Visualize results
Repeat!
21.8 Cloud Computing and Big Data
21.8.1 Scaling Up
The problem:
Genomic datasets are HUGE
Laptop can’t handle it
Need supercomputers
Solution: Cloud computing!
Benefits:
Rent computing power as needed
Scale up or down
Pay only for what you use
No need to buy expensive servers
Platforms:
Amazon Web Services (AWS)
Google Cloud
Microsoft Azure
Specialized: DNAnexus, Seven Bridges
21.9 Workflows and Pipelines
21.9.1 Automating Analysis
Bioinformatics pipeline = Series of analysis steps run automatically
Example RNA-seq pipeline:
Quality control (check raw data)
Trim adapters
Align to genome
Count reads per gene
Normalize
Statistical analysis
Visualize results
Tools for building pipelines:
Nextflow: Modern, powerful
Snakemake: Python-based
Galaxy: Web-based (no programming!)
WDL: Workflow Description Language
Benefits:
Reproducible
Automated
Scalable
Shareable
21.10 Challenges in Bioinformatics
21.10.1 Current Problems
1. Data Quality:
Garbage in, garbage out
Sequencing errors
Sample contamination
Need better quality control
2. Data Integration:
Combining different data types
Genomics + transcriptomics + proteomics
Different formats, scales, biases
Multi-omics challenge!
3. Interpretation:
Finding genes is easy
Understanding function is hard
Most genes poorly characterized
4. Reproducibility:
Different tools give different results
Version control important
Need standard pipelines
5. Computational Resources:
Always need more!
Costs can be high
Environmental impact of computing
21.11 Career Paths in Bioinformatics
21.11.1 Exciting Opportunities!
What bioinformaticians do:
Develop new algorithms
Analyze genomic data
Build databases
Create visualization tools
Apply ML to biology
Collaborate with biologists
Where they work:
Universities (research)
Pharmaceutical companies (drug discovery)
Biotech startups (diagnostics)
Hospitals (clinical genomics)
Government (public health)
Tech companies (Google, Amazon, Microsoft)
Skills needed:
Biology knowledge
Programming (Python, R)
Statistics
Problem-solving
Communication (work with biologists!)
High demand:
More data than people to analyze it
Great job prospects
Competitive salaries
21.12 The Future of Bioinformatics
21.12.1 What’s Coming
1. Real-Time Analysis:
Analyze as data is generated
Feedback during experiments
Faster discoveries
2. AI Integration:
AI assistants for data analysis
Automated interpretation
Hypothesis generation
3. Personalized Medicine:
Analyze your genome on your phone
Instant health insights
Continuous monitoring
4. Synthetic Biology Design:
Design organisms on computer
Predict behavior before building
Engineering life
5. Multi-Omics Integration:
Combine all data types
Complete cell picture
Systems biology realization
21.13 Key Takeaways
Bioinformatics uses computers to analyze biological data
Essential due to massive data from modern biology
Core tasks: Sequence analysis, assembly, annotation, variant calling
Databases store and share biological information
BLAST is the Google of genomics
AlphaFold revolutionized protein structure prediction with AI
Machine learning increasingly important in biology
Programming (Python, R) is essential skill
Cloud computing handles big data
Pipelines automate and standardize analyses
High demand for bioinformatics skills
Future bright with AI integration and personalized medicine
Sources: Information adapted from bioinformatics textbooks, NCBI resources, and computational biology literature.
21.13.1 Best Practices for Primer Design
Do’s: ✅ Use online tools (Primer3, Primer-BLAST) ✅ Check specificity with BLAST ✅ Design primers with similar Tm ✅ Avoid secondary structures ✅ For qPCR: span exon junctions ✅ Include positive and negative controls ✅ Validate experimentally
Don’ts: ❌ Don’t use primers with >3 G/C at 3’ end ❌ Don’t ignore secondary structure prediction ❌ Don’t skip BLAST specificity check ❌ Don’t design primers in repetitive regions ❌ Don’t use primers from different species (unless checking conservation)
21.14 Structural Bioinformatics
21.14.1 Predicting Protein Structures
Why it matters:
Structure determines function
Drug design needs structures
Understanding disease mutations
The problem:
Experiments are slow and expensive
Can we predict structure from sequence?
21.14.2 AlphaFold: AI Solves 50-Year Problem!
AlphaFold (Google DeepMind, 2020):
Uses deep learning / AI
Predicts 3D structure from sequence
Near-experimental accuracy!
Revolutionary!
Impact:
Predicted structures for 200+ million proteins!
Free database (AlphaFold DB)
Accelerating drug discovery
Nobel Prize potential
21.14.3 Homology Modeling
Approach:
If protein A’s structure is known
And protein B is similar sequence
Model B based on A’s structure
Works well for similar proteins!
21.14.4 Protein-Protein Docking
Predicting how proteins interact:
Important for understanding cell signaling
Drug development
Protein engineering
21.15 Machine Learning in Biology
21.15.1 AI Meets Genomics
Machine learning = Teaching computers to learn patterns from data
Applications:
1. Disease Prediction:
Analyze genomes to predict disease risk
Better than humans at finding subtle patterns
Personalized risk scores
2. Variant Interpretation:
Millions of variants in each genome
Which are disease-causing?
ML helps classify them
3. Drug Discovery:
Predict which molecules bind to targets
Design new drugs
Much faster than traditional methods
4. Image Analysis:
Analyze microscopy images
Count cells automatically
Detect cancer in pathology slides
5. Gene Regulation:
Predict which sequences control genes
Understand regulatory code
Design synthetic promoters
21.15.2 Deep Learning Revolution
Deep learning = Advanced ML using neural networks
Successes:
AlphaFold (protein structure)
DeepVariant (variant calling)
BaseNJ (DNA sequencing accuracy)
Drug response prediction
Future: AI will be essential partner in biology!
21.16 Genomic Data Formats
21.16.1 Standard File Formats
FASTA:
Sequence format
Simple text file
Header
ATCGATCG…
FASTQ:
Sequence + quality scores
Raw sequencing data
SAM/BAM:
Aligned sequences
Maps reads to genome
BAM is compressed SAM
VCF (Variant Call Format):
Lists genetic variants
Position, reference, alternate
Standard for sharing variants
GFF/GTF:
Gene annotations
Where genes are located
Exons, introns, etc.
21.16.2 Why Standards Matter
Benefits:
Different tools work together
Share data easily
Reproducible research
Collaborate globally
21.17 Programming in Bioinformatics
21.17.1 Common Languages
Python:
Most popular in bioinformatics
Easy to learn
Powerful libraries (BioPython)
Great for data analysis
R:
Statistical computing
Excellent for genomics
BioConductor (huge package collection)
Beautiful visualizations
Perl:
Text processing
Older but still used
BioPerl
Unix/Linux Command Line:
Essential skill!
File manipulation
Running tools
Automating workflows
Typical workflow:
Process data with Unix commands
Analyze with Python/R
Visualize results
Repeat!
21.18 Cloud Computing and Big Data
21.18.1 Scaling Up
The problem:
Genomic datasets are HUGE
Laptop can’t handle it
Need supercomputers
Solution: Cloud computing!
Benefits:
Rent computing power as needed
Scale up or down
Pay only for what you use
No need to buy expensive servers
Platforms:
Amazon Web Services (AWS)
Google Cloud
Microsoft Azure
Specialized: DNAnexus, Seven Bridges
21.19 Workflows and Pipelines
21.19.1 Automating Analysis
Bioinformatics pipeline = Series of analysis steps run automatically
Example RNA-seq pipeline:
Quality control (check raw data)
Trim adapters
Align to genome
Count reads per gene
Normalize
Statistical analysis
Visualize results
Tools for building pipelines:
Nextflow: Modern, powerful
Snakemake: Python-based
Galaxy: Web-based (no programming!)
WDL: Workflow Description Language
Benefits:
Reproducible
Automated
Scalable
Shareable
21.20 Challenges in Bioinformatics
21.20.1 Current Problems
1. Data Quality:
Garbage in, garbage out
Sequencing errors
Sample contamination
Need better quality control
2. Data Integration:
Combining different data types
Genomics + transcriptomics + proteomics
Different formats, scales, biases
Multi-omics challenge!
3. Interpretation:
Finding genes is easy
Understanding function is hard
Most genes poorly characterized
4. Reproducibility:
Different tools give different results
Version control important
Need standard pipelines
5. Computational Resources:
Always need more!
Costs can be high
Environmental impact of computing
21.21 Career Paths in Bioinformatics
21.21.1 Exciting Opportunities!
What bioinformaticians do:
Develop new algorithms
Analyze genomic data
Build databases
Create visualization tools
Apply ML to biology
Collaborate with biologists
Where they work:
Universities (research)
Pharmaceutical companies (drug discovery)
Biotech startups (diagnostics)
Hospitals (clinical genomics)
Government (public health)
Tech companies (Google, Amazon, Microsoft)
Skills needed:
Biology knowledge
Programming (Python, R)
Statistics
Problem-solving
Communication (work with biologists!)
High demand:
More data than people to analyze it
Great job prospects
Competitive salaries
21.22 The Future of Bioinformatics
21.22.1 What’s Coming
1. Real-Time Analysis:
Analyze as data is generated
Feedback during experiments
Faster discoveries
2. AI Integration:
AI assistants for data analysis
Automated interpretation
Hypothesis generation
3. Personalized Medicine:
Analyze your genome on your phone
Instant health insights
Continuous monitoring
4. Synthetic Biology Design:
Design organisms on computer
Predict behavior before building
Engineering life
5. Multi-Omics Integration:
Combine all data types
Complete cell picture
Systems biology realization
21.23 Key Takeaways
Bioinformatics uses computers to analyze biological data
Essential due to massive data from modern biology
Core tasks: Sequence analysis, assembly, annotation, variant calling
Databases store and share biological information
BLAST is the Google of genomics
AlphaFold revolutionized protein structure prediction with AI
Machine learning increasingly important in biology
Programming (Python, R) is essential skill
Cloud computing handles big data
Pipelines automate and standardize analyses
High demand for bioinformatics skills
Future bright with AI integration and personalized medicine
Sources: Information adapted from bioinformatics textbooks, NCBI resources, and computational biology literature.
21.23.1 Best Practices for Primer Design
Do’s: ✅ Use online tools (Primer3, Primer-BLAST) ✅ Check specificity with BLAST ✅ Design primers with similar Tm ✅ Avoid secondary structures ✅ For qPCR: span exon junctions ✅ Include positive and negative controls ✅ Validate experimentally
Don’ts: ❌ Don’t use primers with >3 G/C at 3’ end ❌ Don’t ignore secondary structure prediction ❌ Don’t skip BLAST specificity check ❌ Don’t design primers in repetitive regions ❌ Don’t use primers from different species (unless checking conservation)
21.24 Structural Bioinformatics
21.24.1 Predicting Protein Structures
Why it matters:
Structure determines function
Drug design needs structures
Understanding disease mutations
The problem:
Experiments are slow and expensive
Can we predict structure from sequence?
21.24.2 AlphaFold: AI Solves 50-Year Problem!
AlphaFold (Google DeepMind, 2020):
Uses deep learning / AI
Predicts 3D structure from sequence
Near-experimental accuracy!
Revolutionary!
Impact:
Predicted structures for 200+ million proteins!
Free database (AlphaFold DB)
Accelerating drug discovery
Nobel Prize potential
21.24.3 Homology Modeling
Approach:
If protein A’s structure is known
And protein B is similar sequence
Model B based on A’s structure
Works well for similar proteins!
21.24.4 Protein-Protein Docking
Predicting how proteins interact:
Important for understanding cell signaling
Drug development
Protein engineering
21.25 Machine Learning in Biology
21.25.1 AI Meets Genomics
Machine learning = Teaching computers to learn patterns from data
Applications:
1. Disease Prediction:
Analyze genomes to predict disease risk
Better than humans at finding subtle patterns
Personalized risk scores
2. Variant Interpretation:
Millions of variants in each genome
Which are disease-causing?
ML helps classify them
3. Drug Discovery:
Predict which molecules bind to targets
Design new drugs
Much faster than traditional methods
4. Image Analysis:
Analyze microscopy images
Count cells automatically
Detect cancer in pathology slides
5. Gene Regulation:
Predict which sequences control genes
Understand regulatory code
Design synthetic promoters
21.25.2 Deep Learning Revolution
Deep learning = Advanced ML using neural networks
Successes:
AlphaFold (protein structure)
DeepVariant (variant calling)
BaseNJ (DNA sequencing accuracy)
Drug response prediction
Future: AI will be essential partner in biology!
21.26 Genomic Data Formats
21.26.1 Standard File Formats
FASTA:
Sequence format
Simple text file
Header
ATCGATCG…
FASTQ:
Sequence + quality scores
Raw sequencing data
SAM/BAM:
Aligned sequences
Maps reads to genome
BAM is compressed SAM
VCF (Variant Call Format):
Lists genetic variants
Position, reference, alternate
Standard for sharing variants
GFF/GTF:
Gene annotations
Where genes are located
Exons, introns, etc.
21.26.2 Why Standards Matter
Benefits:
Different tools work together
Share data easily
Reproducible research
Collaborate globally
21.27 Programming in Bioinformatics
21.27.1 Common Languages
Python:
Most popular in bioinformatics
Easy to learn
Powerful libraries (BioPython)
Great for data analysis
R:
Statistical computing
Excellent for genomics
BioConductor (huge package collection)
Beautiful visualizations
Perl:
Text processing
Older but still used
BioPerl
Unix/Linux Command Line:
Essential skill!
File manipulation
Running tools
Automating workflows
Typical workflow:
Process data with Unix commands
Analyze with Python/R
Visualize results
Repeat!
21.28 Cloud Computing and Big Data
21.28.1 Scaling Up
The problem:
Genomic datasets are HUGE
Laptop can’t handle it
Need supercomputers
Solution: Cloud computing!
Benefits:
Rent computing power as needed
Scale up or down
Pay only for what you use
No need to buy expensive servers
Platforms:
Amazon Web Services (AWS)
Google Cloud
Microsoft Azure
Specialized: DNAnexus, Seven Bridges
21.29 Workflows and Pipelines
21.29.1 Automating Analysis
Bioinformatics pipeline = Series of analysis steps run automatically
Example RNA-seq pipeline:
Quality control (check raw data)
Trim adapters
Align to genome
Count reads per gene
Normalize
Statistical analysis
Visualize results
Tools for building pipelines:
Nextflow: Modern, powerful
Snakemake: Python-based
Galaxy: Web-based (no programming!)
WDL: Workflow Description Language
Benefits:
Reproducible
Automated
Scalable
Shareable
21.30 Challenges in Bioinformatics
21.30.1 Current Problems
1. Data Quality:
Garbage in, garbage out
Sequencing errors
Sample contamination
Need better quality control
2. Data Integration:
Combining different data types
Genomics + transcriptomics + proteomics
Different formats, scales, biases
Multi-omics challenge!
3. Interpretation:
Finding genes is easy
Understanding function is hard
Most genes poorly characterized
4. Reproducibility:
Different tools give different results
Version control important
Need standard pipelines
5. Computational Resources:
Always need more!
Costs can be high
Environmental impact of computing
21.31 Career Paths in Bioinformatics
21.31.1 Exciting Opportunities!
What bioinformaticians do:
Develop new algorithms
Analyze genomic data
Build databases
Create visualization tools
Apply ML to biology
Collaborate with biologists
Where they work:
Universities (research)
Pharmaceutical companies (drug discovery)
Biotech startups (diagnostics)
Hospitals (clinical genomics)
Government (public health)
Tech companies (Google, Amazon, Microsoft)
Skills needed:
Biology knowledge
Programming (Python, R)
Statistics
Problem-solving
Communication (work with biologists!)
High demand:
More data than people to analyze it
Great job prospects
Competitive salaries
21.32 The Future of Bioinformatics
21.32.1 What’s Coming
1. Real-Time Analysis:
Analyze as data is generated
Feedback during experiments
Faster discoveries
2. AI Integration:
AI assistants for data analysis
Automated interpretation
Hypothesis generation
3. Personalized Medicine:
Analyze your genome on your phone
Instant health insights
Continuous monitoring
4. Synthetic Biology Design:
Design organisms on computer
Predict behavior before building
Engineering life
5. Multi-Omics Integration:
Combine all data types
Complete cell picture
Systems biology realization
21.33 Key Takeaways
Bioinformatics uses computers to analyze biological data
Essential due to massive data from modern biology
Core tasks: Sequence analysis, assembly, annotation, variant calling
Databases store and share biological information
BLAST is the Google of genomics
AlphaFold revolutionized protein structure prediction with AI
Machine learning increasingly important in biology
Programming (Python, R) is essential skill
Cloud computing handles big data
Pipelines automate and standardize analyses
High demand for bioinformatics skills
Future bright with AI integration and personalized medicine
Sources: Information adapted from bioinformatics textbooks, NCBI resources, and computational biology literature.
21.33.1 Best Practices for Primer Design
Do’s: ✅ Use online tools (Primer3, Primer-BLAST) ✅ Check specificity with BLAST ✅ Design primers with similar Tm ✅ Avoid secondary structures ✅ For qPCR: span exon junctions ✅ Include positive and negative controls ✅ Validate experimentally
Don’ts: ❌ Don’t use primers with >3 G/C at 3’ end ❌ Don’t ignore secondary structure prediction ❌ Don’t skip BLAST specificity check ❌ Don’t design primers in repetitive regions ❌ Don’t use primers from different species (unless checking conservation)
21.34 Structural Bioinformatics
21.34.1 Predicting Protein Structures
Why it matters:
Structure determines function
Drug design needs structures
Understanding disease mutations
The problem:
Experiments are slow and expensive
Can we predict structure from sequence?
21.34.2 AlphaFold: AI Solves 50-Year Problem!
AlphaFold (Google DeepMind, 2020):
Uses deep learning / AI
Predicts 3D structure from sequence
Near-experimental accuracy!
Revolutionary!
Impact:
Predicted structures for 200+ million proteins!
Free database (AlphaFold DB)
Accelerating drug discovery
Nobel Prize potential
21.34.3 Homology Modeling
Approach:
If protein A’s structure is known
And protein B is similar sequence
Model B based on A’s structure
Works well for similar proteins!
21.34.4 Protein-Protein Docking
Predicting how proteins interact:
Important for understanding cell signaling
Drug development
Protein engineering
21.35 Machine Learning in Biology
21.35.1 AI Meets Genomics
Machine learning = Teaching computers to learn patterns from data
Applications:
1. Disease Prediction:
Analyze genomes to predict disease risk
Better than humans at finding subtle patterns
Personalized risk scores
2. Variant Interpretation:
Millions of variants in each genome
Which are disease-causing?
ML helps classify them
3. Drug Discovery:
Predict which molecules bind to targets
Design new drugs
Much faster than traditional methods
4. Image Analysis:
Analyze microscopy images
Count cells automatically
Detect cancer in pathology slides
5. Gene Regulation:
Predict which sequences control genes
Understand regulatory code
Design synthetic promoters
21.35.2 Deep Learning Revolution
Deep learning = Advanced ML using neural networks
Successes:
AlphaFold (protein structure)
DeepVariant (variant calling)
BaseNJ (DNA sequencing accuracy)
Drug response prediction
Future: AI will be essential partner in biology!
21.36 Genomic Data Formats
21.36.1 Standard File Formats
FASTA:
Sequence format
Simple text file
Header
ATCGATCG…
FASTQ:
Sequence + quality scores
Raw sequencing data
SAM/BAM:
Aligned sequences
Maps reads to genome
BAM is compressed SAM
VCF (Variant Call Format):
Lists genetic variants
Position, reference, alternate
Standard for sharing variants
GFF/GTF:
Gene annotations
Where genes are located
Exons, introns, etc.
21.36.2 Why Standards Matter
Benefits:
Different tools work together
Share data easily
Reproducible research
Collaborate globally
21.37 Programming in Bioinformatics
21.37.1 Common Languages
Python:
Most popular in bioinformatics
Easy to learn
Powerful libraries (BioPython)
Great for data analysis
R:
Statistical computing
Excellent for genomics
BioConductor (huge package collection)
Beautiful visualizations
Perl:
Text processing
Older but still used
BioPerl
Unix/Linux Command Line:
Essential skill!
File manipulation
Running tools
Automating workflows
Typical workflow:
Process data with Unix commands
Analyze with Python/R
Visualize results
Repeat!
21.38 Cloud Computing and Big Data
21.38.1 Scaling Up
The problem:
Genomic datasets are HUGE
Laptop can’t handle it
Need supercomputers
Solution: Cloud computing!
Benefits:
Rent computing power as needed
Scale up or down
Pay only for what you use
No need to buy expensive servers
Platforms:
Amazon Web Services (AWS)
Google Cloud
Microsoft Azure
Specialized: DNAnexus, Seven Bridges
21.39 Workflows and Pipelines
21.39.1 Automating Analysis
Bioinformatics pipeline = Series of analysis steps run automatically
Example RNA-seq pipeline:
Quality control (check raw data)
Trim adapters
Align to genome
Count reads per gene
Normalize
Statistical analysis
Visualize results
Tools for building pipelines:
Nextflow: Modern, powerful
Snakemake: Python-based
Galaxy: Web-based (no programming!)
WDL: Workflow Description Language
Benefits:
Reproducible
Automated
Scalable
Shareable
21.40 Challenges in Bioinformatics
21.40.1 Current Problems
1. Data Quality:
Garbage in, garbage out
Sequencing errors
Sample contamination
Need better quality control
2. Data Integration:
Combining different data types
Genomics + transcriptomics + proteomics
Different formats, scales, biases
Multi-omics challenge!
3. Interpretation:
Finding genes is easy
Understanding function is hard
Most genes poorly characterized
4. Reproducibility:
Different tools give different results
Version control important
Need standard pipelines
5. Computational Resources:
Always need more!
Costs can be high
Environmental impact of computing
21.41 Career Paths in Bioinformatics
21.41.1 Exciting Opportunities!
What bioinformaticians do:
Develop new algorithms
Analyze genomic data
Build databases
Create visualization tools
Apply ML to biology
Collaborate with biologists
Where they work:
Universities (research)
Pharmaceutical companies (drug discovery)
Biotech startups (diagnostics)
Hospitals (clinical genomics)
Government (public health)
Tech companies (Google, Amazon, Microsoft)
Skills needed:
Biology knowledge
Programming (Python, R)
Statistics
Problem-solving
Communication (work with biologists!)
High demand:
More data than people to analyze it
Great job prospects
Competitive salaries
21.42 The Future of Bioinformatics
21.42.1 What’s Coming
1. Real-Time Analysis:
Analyze as data is generated
Feedback during experiments
Faster discoveries
2. AI Integration:
AI assistants for data analysis
Automated interpretation
Hypothesis generation
3. Personalized Medicine:
Analyze your genome on your phone
Instant health insights
Continuous monitoring
4. Synthetic Biology Design:
Design organisms on computer
Predict behavior before building
Engineering life
5. Multi-Omics Integration:
Combine all data types
Complete cell picture
Systems biology realization
21.43 Key Takeaways
Bioinformatics uses computers to analyze biological data
Essential due to massive data from modern biology
Core tasks: Sequence analysis, assembly, annotation, variant calling
Databases store and share biological information
BLAST is the Google of genomics
AlphaFold revolutionized protein structure prediction with AI
Machine learning increasingly important in biology
Programming (Python, R) is essential skill
Cloud computing handles big data
Pipelines automate and standardize analyses
High demand for bioinformatics skills
Future bright with AI integration and personalized medicine
Sources: Information adapted from bioinformatics textbooks, NCBI resources, and computational biology literature.