Genomics Knowledge Base
A free, comprehensive learning guide covering Whole Exome Sequencing (WES), NGS pipelines, ACMG variant classification, FASTQ/BAM/VCF formats, variant annotation, RNA-seq, single-cell genomics, and clinical genomics pipelines - from beginner to advanced.
What You Will Learn
Introduction to Genomics
Genomics studies the complete genetic material of an organism. It combines molecular biology, genetics, statistics, and computer science. Clinical genomics has transformed rare disease diagnosis - whole exome sequencing provides diagnoses for 25–35% of previously undiagnosed patients. Real-world example: a 4-year-old with 3 years of inconclusive investigations was diagnosed with HSD10 mitochondrial disease by WES trio analysis in 6 weeks, changing clinical management and enabling accurate genetic counselling for the family.
Fundamental Biology
DNA's double helix pairs A with T (2 hydrogen bonds) and C with G (3 bonds). The central dogma: DNA → pre-mRNA → mRNA → Protein. Each step offers a disease mechanism: nonsense variants create premature stop codons (PVS1 in ACMG), splice-site variants disrupt GT...AG splice sites, and missense variants change amino acids. Example: BRCA1 c.5266dupC creates a frameshift p.(Gln1756Profs*74) - a pathogenic Ashkenazi Jewish founder mutation causing hereditary breast cancer.
Genomics Technologies
Sanger sequencing is the traditional gold-standard method with read lengths of 600–1000 bp and >99.9% accuracy. Next Generation Sequencing (NGS) enables high-throughput sequencing of millions of reads simultaneously. Major platforms include Illumina (short reads, high accuracy, most widely used), PacBio (long reads, moderate accuracy), Oxford Nanopore (real-time long reads), and Ion Torrent (semiconductor-based).
Types of Sequencing
Whole Genome Sequencing (WGS) sequences the entire genome including coding and non-coding regions. Whole Exome Sequencing (WES) captures only the coding regions (~1–2% of genome), making it cost-effective for disease studies and rare variant discovery. RNA Sequencing (RNA-seq) measures gene expression levels. Targeted sequencing focuses on selected gene panels at high depth. Single-cell sequencing enables cell-level molecular profiling.
Linux for Genomics
Essential Linux commands for genomics include: pwd (current directory), ls -lh (list files with sizes), mkdir (create directory), cp/mv (copy/move files), cat/head/tail (view file contents), grep (search patterns), awk and sed for text processing, and find for file searching. For FASTQ files: zcat allows viewing compressed files, and wc -l counts lines (divide by 4 for read count).
Common File Formats
Key genomics file formats: FASTQ (.fastq/.fq.gz) stores raw sequencing reads with quality scores. FASTA (.fa/.fasta) stores reference sequences. SAM/BAM (.sam/.bam) stores aligned reads. VCF (.vcf/.vcf.gz) stores variant calls. BED files store genomic coordinates. GTF/GFF files store gene annotations. Phred quality score Q = -10 × log₁₀(P): Q30 = 99.9% accuracy, Q20 = 99% accuracy.
NGS Workflow Overview
The standard NGS bioinformatics pipeline involves: Sample Collection → DNA Extraction → Library Preparation → Sequencing → FASTQ Generation → Quality Control → Alignment to reference genome → Variant Calling → Annotation → Clinical Interpretation. Each step requires specific tools and quality checks to ensure accurate results.
FASTQ Quality Control
FastQC is the most widely-used QC tool for raw sequencing reads. Key metrics include per-base sequence quality (target Q30+), per-sequence quality scores, GC content (~50% expected), adapter content (<5% preferred), and sequence duplication levels (<20% recommended). MultiQC aggregates multiple QC reports into a single interactive HTML report.
Read Trimming
Trimmomatic performs paired-end trimming with adapter removal (ILLUMINACLIP), quality-based trimming (LEADING, TRAILING, SLIDINGWINDOW), and minimum length filtering (MINLEN). fastp is a modern all-in-one QC and trimming tool that is faster and automatically detects adapters for paired-end data, outputting both JSON and HTML reports.
Sequence Alignment
Reference genomes GRCh37/hg19 (older, widely used) and GRCh38/hg38 (current standard, recommended). BWA-MEM is the standard aligner for short reads: index the reference with bwa index, align with bwa mem using read group tags (-R flag), convert to BAM with samtools view, sort with samtools sort, and index with samtools index. Check alignment statistics with samtools flagstat.
Post-Alignment Processing
Picard MarkDuplicates identifies PCR and optical duplicates in aligned BAM files. Base Quality Score Recalibration (BQSR) corrects systematic errors using machine learning: run GATK BaseRecalibrator to generate a recalibration table using known variant sites (dbSNP, Mills, 1000G), then apply with GATK ApplyBQSR. These steps are mandatory for variant calling accuracy.
Variant Calling
GATK HaplotypeCaller is the gold standard for germline variant calling. Run in GVCF mode per sample, then use GenotypeGVCFs for joint calling across multiple samples. For somatic variants (cancer), use GATK Mutect2 with paired tumor-normal BAM files. Variant Quality Score Recalibration (VQSR) further filters variants using machine learning trained on known sites.
Understanding VCF Files
VCF (Variant Call Format) files store genetic variants. Each variant row contains: CHROM (chromosome), POS (1-based position), ID (rs number), REF (reference allele), ALT (alternate allele), QUAL (quality score), FILTER (PASS or filter reason), INFO (metadata like depth and allele frequency), FORMAT (field definitions), and SAMPLE (genotype: 0/0=homozygous reference, 0/1=heterozygous, 1/1=homozygous alternate).
Variant Annotation
ANNOVAR annotates variants using multiple databases: convert VCF to ANNOVAR input format with convert2annovar.pl, then run table_annovar.pl with protocols for refGene (functional effect), ClinVar (clinical significance), gnomAD (population frequency), dbNSFP (functional prediction scores including SIFT, PolyPhen, CADD), and COSMIC (cancer mutations). VEP (Variant Effect Predictor) from Ensembl provides alternative comprehensive annotation.
Population Databases
Key genomic databases: gnomAD contains allele frequencies from >800,000 exomes and genomes across diverse populations. ClinVar is the authoritative source for clinical variant classifications. 1000 Genomes Project provides global population variation data. dbSNP is the comprehensive SNP repository. ExAC is the Exome Aggregation Consortium database. HGMD (Human Gene Mutation Database) catalogs disease-causing mutations.
ACMG Variant Interpretation
ACMG/AMP 2015 guidelines classify variants into 5 categories: Pathogenic (P), Likely Pathogenic (LP), Variant of Uncertain Significance (VUS), Likely Benign (LB), and Benign (B). Evidence criteria include: PVS1 (null variant where LOF is disease mechanism), PS1-PS4 (strong pathogenic), PM1-PM6 (moderate pathogenic), PP1-PP5 (supporting pathogenic), BA1 (allele frequency >5% = benign), and BS1-BS4 (strong benign evidence).
RNA-seq Analysis
RNA-seq workflow: FASTQ → QC → Alignment (STAR or HISAT2) → Count Matrix (featureCounts or HTSeq) → Differential Expression → Pathway Analysis. STAR aligner is fast and splice-aware. DESeq2 (R) is the gold standard for differential expression: create DESeqDataSetFromMatrix, run DESeq(), extract results, and filter by padj < 0.05 and |log2FoldChange| > 1.
Single-cell Genomics
Single-cell RNA sequencing reveals cell-level transcriptomic diversity. Key analysis steps: quality control (filter low-quality cells), normalization, dimensionality reduction (PCA, UMAP, t-SNE), clustering, and cell type annotation. Popular tools: Seurat (R, most widely used), Scanpy (Python, scalable), Cell Ranger (10X Genomics preprocessing), Monocle (trajectory analysis), and Harmony (batch correction).
Structural Variant Analysis
Structural variants (SVs) are genomic rearrangements ≥50bp: deletions (DEL), insertions (INS), inversions (INV), duplications (DUP), and translocations (TRA). Detection tools: DELLY uses paired-end and split-read signals, Manta is fast and clinical-grade, LUMPY uses a probabilistic framework, CNVkit detects copy number variants in WES, and GRIDSS supports long-read data.
Long-read Sequencing
Long-read sequencing (PacBio HiFi, Oxford Nanopore) offers advantages: better structural variant detection, resolution of repetitive regions (centromeres, telomeres), full-length transcript analysis, direct epigenetic modification detection (Nanopore), and haplotype phasing. Key tools: Minimap2 for alignment, Sniffles2 and CuteSV for SV calling, Dorado for Nanopore basecalling.
Programming for Genomics
Essential Python libraries for genomics: Biopython for sequence parsing and BLAST queries, pysam for BAM/SAM file handling, cyvcf2 for fast VCF parsing, pandas for data manipulation, and matplotlib/seaborn for visualization. Key operations: parse FASTA with SeqIO, iterate BAM reads with pysam.AlignmentFile, extract genotypes from VCF with cyvcf2.VCF.
R for Genomics
Bioconductor provides 2000+ packages for genomic analysis in R. Key packages: GenomicRanges for genomic interval operations, DESeq2 and edgeR for differential expression, ggplot2 for publication-quality visualization, VariantAnnotation for VCF handling, BSgenome for reference sequences. Read VCF files with readVcf(), extract genotypes with geno()$GT, and access INFO fields with info().
Cloud and HPC Genomics
Workflow management systems enable reproducible genomics: Nextflow (DSL2, portable across HPC/cloud), Snakemake (Python-based), WDL/Cromwell (Broad/Google standard), and CWL (Common Workflow Language). Containerization ensures reproducibility: Docker packages tools with dependencies, Singularity/Apptainer is HPC-safe, and Conda manages software environments.
Clinical Genomics Pipeline
Clinical genomics workflow: Patient Sample → Sequencing → Quality Control → Variant Calling → Filtering → Annotation → ACMG Interpretation → Clinical Report. Variant prioritization uses: rare variants (MAF <1% in gnomAD), predicted damaging effects (SIFT, PolyPhen, CADD>20), disease-associated genes (OMIM), inheritance pattern matching, and phenotype correlation using HPO terms.
AI in Genomics
Artificial intelligence is transforming genomics: AlphaMissense, CADD, and REVEL predict variant pathogenicity. AlphaFold2 revolutionized protein structure prediction achieving atomic accuracy. ESM-2 is a protein language model trained on 250 million sequences. Enformer predicts gene expression from DNA sequence. AlphaMissense classifies 71 million missense variants. The Nucleotide Transformer is a genomic foundation model.
Knowledge Base
Interactive genomics learning guide - from DNA basics to advanced clinical pipelines. Track your progress and learn at your own pace.
Introduction to Genomics
beginner5 sections
Genomics at a Glance
Core Branches of Genomics
What is Genomics?
Genomics is the study of the complete genetic material (genome) of an organism - every DNA sequence including protein-coding genes, regulatory elements, non-coding RNA, and repetitive regions. It integrates molecular biology, genetics, statistics, and computer science to answer one fundamental question: how does the sequence of DNA give rise to the biology of a living organism?
Unlike classical genetics which studies one gene at a time, genomics looks at the entire genome simultaneously. This systems-level view allows us to discover disease genes, understand evolution, predict drug responses, and diagnose rare conditions that were previously undiagnosable.
Core Branches of Genomics
- Structural genomics - sequencing, assembling, and physically mapping genomes; the foundation for everything else
- Functional genomics - determining what genes do: RNA-seq (expression), ChIP-seq (binding), ATAC-seq (accessibility)
- Comparative genomics - aligning genomes across species to identify conserved (important) regions and study evolution
- Clinical genomics - translating sequencing findings into diagnosis, prognosis, and treatment decisions
- Metagenomics - sequencing microbial communities (gut microbiome, environmental samples) without culturing
- Pharmacogenomics - predicting drug response from genetic variation (e.g., CYP2D6 variants affect codeine metabolism)
- Cancer genomics - identifying somatic driver mutations, fusions, and copy number changes in tumours
The Human Genome - Key Numbers
- ~3.2 billion base pairs (3.2 Gb) per haploid genome - if stretched, DNA from one cell would be ~2 metres long
- ~20,000–25,000 protein-coding genes - only ~1.5% of total DNA; the rest is regulatory, repetitive, or non-coding RNA
- 46 chromosomes (22 autosome pairs + XX or XY sex chromosomes) in diploid somatic cells
- ~99.9% identical between any two humans - variants in the remaining 0.1% (~3 million sites) drive most heritable disease
- Human Genome Project completed April 2003 after 13 years and $3 billion; first time any genome was sequenced
- T2T-CHM13 (2022) filled the remaining 8% gaps including centromeres; first truly complete human genome
- Average gene spans ~27 kb genomic space but encodes only ~1.3 kb of protein-coding sequence
Why Genomics Matters Clinically
The clinical impact of genomics has shifted from research curiosity to routine medical practice. Whole exome sequencing now provides diagnoses for 25–35% of patients with rare undiagnosed diseases who had previously exhausted all other investigations.
- Rare disease diagnosis: 1 in 17 people affected worldwide; 80% have a genetic origin; WES/WGS is now first-line for undiagnosed patients
- Cancer genomics: somatic driver mutations guide targeted therapy - EGFR in lung cancer, HER2 in breast, BRAF in melanoma
- Pharmacogenomics: CYP2D6 poor metabolisers accumulate codeine toxicity; HLA-B*57:01 causes abacavir hypersensitivity
- Newborn screening: 35+ conditions including PKU, congenital hypothyroidism, and SMA detectable from a heel-prick blood spot
- Infectious disease: Nanopore sequencing identified SARS-CoV-2 variants in real time; metagenomics diagnoses encephalitis in hours
- Hereditary risk: BRCA1/2 carriers have 70% lifetime breast cancer risk; Lynch syndrome confers 50% colorectal cancer risk
- Preimplantation genetic testing (PGT): embryo selection before IVF for families with known genetic conditions
Real-World Case Study: The Undiagnosed Disease
A 4-year-old boy presented to a genetics clinic in Mumbai with severe hypotonia (low muscle tone), developmental delay, seizures, and elevated lactate. Over 3 years, he had undergone 47 investigations including MRI, muscle biopsy, metabolic panels, and chromosomal microarray - all inconclusive. His family had spent ₹18 lakhs on diagnostics.
Whole exome sequencing of the trio (child + both parents) was performed. Within 6 weeks, a de novo pathogenic variant was identified: NM_004249.3(HSD17B10):c.388A>G p.(Lys130Glu) in the HSD10 gene on the X chromosome, causing HSD10 mitochondrial disease (also known as 2-methyl-3-hydroxybutyryl-CoA dehydrogenase deficiency). This is an ultra-rare X-linked disorder with fewer than 50 reported cases globally.
- Time to diagnosis: 6 weeks by WES vs 3 years without it
- Variant type: de novo missense - required trio sequencing to identify
- Clinical impact: family counselled on recurrence risk (<1% for de novo), genetic testing of mother confirmed carrier status
- Management changed: started thiamine supplementation and mitochondrial cofactors; seizure medications optimised
- This case demonstrates why clinical genomics is transformative for undiagnosed disease
Fundamental Biology
beginner6 sections
Genomics Technologies
beginner5 sections
Types of Sequencing
beginner5 sections
Linux for Genomics
beginner6 sections
Common File Formats
beginner5 sections
NGS Workflow Overview
beginner5 sections
FASTQ Quality Control
intermediate4 sections
Read Trimming
intermediate5 sections
Sequence Alignment
intermediate5 sections
Post-Alignment Processing
intermediate4 sections
Variant Calling
intermediate4 sections
Understanding VCF Files
intermediate4 sections
Variant Annotation
intermediate5 sections
Population & Clinical Databases
intermediate4 sections
ACMG Variant Interpretation
intermediate7 sections
RNA-seq Analysis
intermediate5 sections
Single-cell Genomics
advanced5 sections
Structural Variant Analysis
advanced4 sections
Long-read Sequencing
advanced5 sections
Python for Genomics
intermediate5 sections
R for Genomics
intermediate4 sections
Cloud & HPC Genomics
advanced5 sections
Clinical Genomics Pipeline
advanced6 sections
AI in Genomics
advanced5 sections
Learning Roadmap
beginner6 sections
Interview Questions
beginner5 sections
Hands-on Projects
beginner4 sections
Best Practices
beginner5 sections