Learning Roadmap

Before touching any sequencing data, build a solid foundation in biology, computing, and statistics. Many people skip this and struggle later because they can run a pipeline but cannot interpret the output.

• Biology: DNA structure, central dogma, gene structure (exons/introns/UTRs), types of variants (SNV/indel/CNV/SV)
• Linux: navigation, file operations, grep/awk/sed, shell scripting, managing background jobs
• File formats: FASTQ (quality scores, Phred encoding), FASTA, SAM/BAM (FLAGS, CIGAR), VCF (genotype fields)
• Python or R basics: variables, loops, functions, data frames, reading/writing tabular files
• Statistics: mean, variance, normal distribution, p-value, multiple testing correction (Bonferroni, FDR/BH)
• Milestone check: can you write a bash script that counts reads in 10 FASTQ files and prints the totals?
• Free resources: Coursera "Genomic Data Science" (Johns Hopkins); "Bioinformatics Algorithms" (Compeau & Pevzner); EMBL-EBI training courses

Run real pipelines on real data. Download a public WES dataset from the NCBI Sequence Read Archive (SRA) and process it end-to-end. NA12878 (HG001) is the perfect training sample - it has a validated truth set for benchmarking your results.

• FASTQ QC: FastQC, MultiQC, fastp - interpret every module; understand what adapter contamination looks like
• Alignment: BWA-MEM2 → samtools sort/index → understand CIGAR strings, read groups, mapping quality
• GATK best practices: MarkDuplicates → BQSR → HaplotypeCaller (GVCF mode) → GenotypeGVCFs → VQSR
• Variant annotation: ANNOVAR or VEP - add gnomAD AF, ClinVar significance, CADD score, gene consequence
• ACMG classification: manually classify 5 variants using the 2015 guidelines; compare to ClinVar classification
• RNA-seq starter: STAR alignment → featureCounts → DESeq2; produce a volcano plot
• Milestone: run NA12878 WES → compare your VCF to GIAB truth set using hap.py; target >99% SNV sensitivity

• Single-cell: Seurat or Scanpy full pipeline on a GEO dataset; cell type annotation; trajectory analysis with Monocle3
• Structural variants: Manta + DELLY ensemble calling; CNVkit for copy number from WES; SV interpretation
• Long reads: Minimap2 for PacBio/Nanopore; HiFi variant calling with DeepVariant; phasing with WhatsHap
• Workflow automation: write a 3-step Nextflow DSL2 pipeline with Docker containers; use -resume for checkpointing
• Cloud computing: run a pipeline on AWS Batch or Google Cloud Life Sciences; understand spot instance cost savings
• Choose a focus: clinical genomics (ACMG, report writing), cancer genomics (somatic calling, mutational signatures), or research (population genetics, multi-omics)
• Milestone: submit a methods section describing your pipeline to a lab meeting or bioRxiv preprint

• Machine learning: train a random forest on ClinVar P/B variants; learn PyTorch for deep learning; understand AlphaMissense architecture
• Population genetics: GWAS quality control pipeline; LD pruning; PCA for ancestry; fine-mapping with FINEMAP
• Clinical genomics mastery: phenotype-driven analysis with Exomiser; ACMG secondary findings reporting; variant re-interpretation
• Multi-omics: integrate WES + RNA-seq + ATAC-seq; eQTL analysis; MOFA+ for factor analysis
• Contribute open source: add a module to nf-core, submit a Bioconductor package, or fix a bug in a popular tool
• Certifications: EMQN Next Generation Sequencing scheme for clinical labs; CAP/CLIA competency testing
• Career paths: clinical molecular geneticist, research bioinformatician, computational genomics scientist, genomics product manager

Priya, an MSc biotechnology graduate from Hyderabad with no prior coding experience, became a clinical bioinformatician at a major diagnostic lab in 18 months:

• Month 1–2: Completed "Introduction to Genomic Technologies" on Coursera; learned Linux basics with "The Linux Command Line" book
• Month 3–4: Downloaded NA12878 SRA data; ran FastQC; struggled with BWA index for 3 days (lesson: RTFM)
• Month 5–7: Completed full GATK WES pipeline; achieved 98.9% SNV sensitivity vs GIAB truth set
• Month 8–10: Annotated variants with ANNOVAR; manually classified 50 ClinVar variants; built a Python ACMG scorer
• Month 11–14: Joined a local genomics lab as a research assistant; analysed 20 clinical WES cases under supervision
• Month 15–18: Built a Nextflow pipeline for the lab; presented at a local genomics conference; received job offer
• Key insight: "I spent 80% of my time on 20% of the tasks. Debugging alignment and understanding VCF format took longer than running the tools themselves."