Introduction to Genomic Data Analysis: Biostatistics and Machine Learning for High-Dimensional Biology

Learn to analyze genomic data the way working computational biologists and biostatisticians actually do — from a public repository accession number to a defensible, reproducible result. This intermediate course assumes statistics (regression, hypothesis testing, basic machine learning) and assumes no biology, teaching exactly the molecular vocabulary needed to read a results table and no more. It opens with the landscape: what each assay physically measures, how every technology lands as the same features-by-samples matrix, and the gallery of computational genomics study types — differential expression, class discovery and class prediction, genome-wide association, epigenomic profiling, single-cell atlases — each matched to the question it answers. It then builds the practical toolbox that everything rests on: file formats (FASTQ, BAM, VCF, BED, GTF) and matrix objects, reference genome builds and the coordinate conventions that silently corrupt projects, the gene-identifier mapping problem, Bioconductor's SummarizedExperiment and Python's AnnData, and reproducible environments and workflow managers. A full chapter is devoted to the public data ecosystem — GEO, ArrayExpress, SRA/ENA, dbGaP, TCGA, GTEx, ENCODE, recount3 and UK Biobank — covering open versus controlled access, MIAME and MINSEQE metadata standards, programmatic retrieval in both languages, and the discipline of auditing a downloaded dataset before trusting it. The statistical core follows the canonical path: microarrays end to end (quality assessment, background correction, quantile normalization, RMA summarization, annotation), gene-wise linear models with empirical Bayes shrinkage, the p-value histogram as the field's most informative diagnostic, and false discovery rate control at the scale of twenty thousand simultaneous tests; then RNA-seq counts with negative binomial models, composition-aware normalization, DESeq2, edgeR and limma-voom; then batch effects, confounding, ComBat and surrogate variable analysis, and the design principles that prevent the problem; then interpretation via over-representation and gene set enrichment analysis, with the background-universe and redundancy traps that ruin most published enrichment. The machine learning chapter covers class discovery (PCA, UMAP, clustering, and a single-cell workflow showcase in Seurat and Scanpy) and class prediction under p much greater than n, built around the field's hardest-won lesson: feature selection outside cross-validation manufactures accuracy from pure noise, and only nested validation and independent cohorts prove anything. The course closes with genetic variation and GWAS — quality control, population stratification, the genome-wide significance threshold, QQ plots and genomic inflation, Manhattan plots and polygenic scores — and a capstone reanalysis of a public dataset from accession to written report. Labs run in both R (Bioconductor, limma, DESeq2, edgeR, sva, clusterProfiler, Seurat) and Python (scanpy, pandas, scikit-learn, pydeseq2), with a synthetic ground-truth suite throughout so every FDR claim, batch correction, and classifier can be scored against known truth. The organizing theme: in genomics the data is wide, the noise is structured, and the discipline that separates a finding from an artifact is statistical, not technological.

  • Level: intermediate
  • Duration: 10 weeks
  • 11 sections
  • Price: $50

Topics: genomics, bioinformatics, Bioconductor, gene expression, microarray, RMA, quantile normalization, probeset annotation, RNA-seq, count data, negative binomial, overdispersion, DESeq2, edgeR, limma, limma-voom, empirical Bayes, moderated t-statistic, design matrix, contrasts, differential expression, log fold change, shrinkage, multiple testing, false discovery rate, Benjamini-Hochberg, q-value, local FDR, p-value histogram, volcano plot, MA plot, batch effects, ComBat, surrogate variable analysis, RUV, confounding, experimental design, technical replication, GEO, SRA, ArrayExpress, dbGaP, TCGA, GTEx, ENCODE, recount3, MIAME, MINSEQE, FAIR principles, controlled access, FASTQ, BAM, VCF, BED, GTF, reference genome, GRCh38, liftover, genome coordinates, gene identifiers, SummarizedExperiment, GRanges, AnnData, reproducible research, Snakemake, Nextflow, containers, gene set enrichment analysis, GSEA, over-representation analysis, Gene Ontology, KEGG, Reactome, MSigDB, background universe, pathway analysis, dimension reduction, PCA, UMAP, t-SNE, hierarchical clustering, consensus clustering, single-cell RNA-seq, Seurat, Scanpy, Leiden clustering, class discovery, class prediction, p greater than n, regularization, lasso, elastic net, random forest, selection bias, nested cross-validation, external validation, AUC, GWAS, SNP, minor allele frequency, linkage disequilibrium, Hardy-Weinberg equilibrium, PLINK, population stratification, genomic inflation factor, QQ plot, Manhattan plot, genome-wide significance, polygenic risk score, heritability

What's inside

  1. Introduction to Genomic Data Analysis: Biostatistics and Machine Learning for High-Dimensional Biology
  2. Chapter 1: The Genomic Data Landscape: What Is Measured, Why It Is Hard, and Which Study Is Being Run
  3. Chapter 2: The Genomic Toolbox: Formats, Coordinates, Identifiers, and Reproducible Workflows
  4. Chapter 3: Public Repositories: Finding, Retrieving, and Trusting Genomic Data
  5. Chapter 4: Microarrays: The Canonical Genomics Experiment, End to End
  6. Chapter 5: Differential Expression I: Linear Models, Empirical Bayes, and Multiple Testing
  7. Chapter 6: RNA-seq: Counts, Negative Binomial Models, and Modern Differential Expression
  8. Chapter 7: Batch Effects, Confounding, and Experimental Design
  9. Chapter 8: From Gene Lists to Biology
  10. Chapter 9: Machine Learning on Genomic Data
  11. Chapter 10: Genetic Variation and GWAS — with the Capstone

Read a free sample chapter