Description

This curriculum spans the full workflow of gene expression analysis in bioinformatics, comparable to the technical depth and decision-making structure of a multi-phase research initiative integrating experimental design, multi-omics data processing, and reproducible analysis frameworks used in academic consortia and pharmaceutical discovery programs.

Module 1: Foundations of Gene Expression Technologies

Select RNA-seq over microarrays when detecting novel transcripts or requiring a broader dynamic range in expression quantification.
Choose stranded RNA-seq protocols to resolve antisense transcription and overlapping gene annotations in complex genomes.
Decide between bulk and single-cell RNA-seq based on biological question—tissue heterogeneity versus population-level expression trends.
Implement spike-in controls (e.g., ERCC) to normalize technical variation in low-input or degraded RNA samples.
Evaluate library preparation kits for compatibility with degraded samples (e.g., FFPE tissues) and sequencing platform constraints.
Design multiplexed sequencing runs balancing sample throughput, read depth, and cost per sample.
Establish minimum read depth thresholds (e.g., 20–30 million reads) based on transcriptome complexity and detection sensitivity needs.
Document batch information during sample processing to enable downstream batch correction in analysis.

Module 2: Experimental Design and Sample Quality Control

Randomize sample processing order to minimize batch effects confounded with biological conditions.
Set RIN (RNA Integrity Number) thresholds (e.g., ≥7) for inclusion in downstream analysis, excluding degraded samples.
Include biological replicates (minimum n=3 per condition) to enable statistical power for differential expression detection.
Integrate negative controls (e.g., no-template RT controls) to monitor contamination in library prep.
Balance cohort composition across covariates (e.g., age, sex, batch) to avoid confounding in analysis.
Use PCA on preliminary expression data to identify outliers prior to formal analysis.
Define exclusion criteria for samples based on low alignment rates or high ribosomal RNA content.
Implement blinding during sample processing to reduce operator bias in handling.

Module 3: Raw Data Processing and Alignment

Select alignment tools (e.g., STAR vs. HISAT2) based on speed, memory footprint, and splice junction sensitivity.
Build custom genome indices when working with non-reference strains or engineered organisms.
Trim adapter sequences and low-quality bases using tools like Trimmomatic or Cutadapt before alignment.
Assess alignment metrics (e.g., % uniquely mapped reads, splice junctions detected) for quality assurance.
Handle multimapping reads based on study goals—exclude for gene-level counts or resolve with probabilistic methods.
Filter ribosomal RNA alignments using pre-mapping or post-alignment subtraction with reference databases.
Standardize file formats (e.g., BAM, CRAM) and indexing for efficient data access and sharing.
Validate alignment reproducibility across replicates using correlation of coverage profiles.

Module 4: Quantification and Normalization Strategies

Choose featureCounts or HTSeq for gene-level counts when prioritizing simplicity and compatibility with DE tools.
Use transcript-level quantifiers (e.g., Salmon, kallisto) with alignment-free methods for improved isoform resolution.
Apply TMM normalization in edgeR for library size and composition bias correction in differential expression.
Compare normalization methods (e.g., TPM, FPKM, DESeq2’s median-of-ratios) based on downstream use case.
Adjust for gene length and GC content when comparing expression across genes or studies.
Retain raw counts for statistical testing, avoiding pre-normalized data that limits reanalysis options.
Account for sequencing depth differences when integrating datasets from multiple batches or studies.
Monitor the impact of normalization on variance structure using PCA before differential expression analysis.

Module 5: Differential Expression Analysis

Select DESeq2, edgeR, or limma-voom based on data distribution, sample size, and design complexity.
Model batch effects as covariates in the design matrix to prevent false positives.
Set significance thresholds using adjusted p-values (e.g., FDR < 0.05) and minimum log2 fold change (e.g., |1.0|).
Validate dispersion estimates and mean-variance trends to ensure model fit in count-based methods.
Perform power analysis post-hoc to interpret non-significant results in underpowered studies.
Use contrasts to test specific hypotheses (e.g., time-point comparisons, interaction effects).
Generate MA and volcano plots with gene labels to communicate results to domain experts.
Export ranked gene lists for pathway enrichment and prioritization in validation experiments.

Module 6: Functional Enrichment and Pathway Analysis

Map gene identifiers consistently across databases (e.g., Ensembl, Entrez, HGNC) to avoid annotation mismatches.
Select background gene sets that reflect detection capability (e.g., expressed genes) rather than the whole genome.
Compare over-representation analysis (ORA) with gene set enrichment analysis (GSEA) based on data continuity.
Adjust for gene length bias in enrichment results when using RNA-seq data with positional biases.
Use curated pathway databases (e.g., Reactome, MSigDB) with version-controlled annotations.
Interpret enrichment results in context of tissue-specific expression and known biological roles.
Validate enrichment findings with orthogonal data (e.g., protein levels, phenotypic assays).
Report multiple testing correction methods applied to enrichment p-values (e.g., FDR, Bonferroni).

Module 7: Single-Cell RNA-Seq Analysis Pipeline

Set UMI and cell barcode thresholds to distinguish real cells from ambient RNA and empty droplets.
Apply doublet detection algorithms (e.g., Scrublet, DoubletFinder) in droplet-based scRNA-seq data.
Select dimensionality reduction methods (PCA, UMAP, t-SNE) based on interpretability and computational load.
Choose clustering resolution parameters to balance granularity and biological coherence.
Annotate cell types using marker genes from reference atlases or literature, avoiding over-clustering artifacts.
Correct batch effects across samples using integration methods (e.g., Harmony, Seurat’s CCA) without removing biological variation.
Filter mitochondrial gene percentage and total UMI counts to remove low-quality or stressed cells.
Validate pseudotime inference results with known differentiation markers and trajectory topology.

Module 8: Data Integration and Multi-Omics Considerations

Align genomic coordinates and gene annotations across data types (e.g., RNA-seq, ChIP-seq, methylation) using consistent reference builds.
Use WGCNA or MOFA+ to identify co-expression modules correlated with epigenetic or clinical traits.
Match sample IDs rigorously across omics layers, resolving discrepancies in naming or processing dates.
Normalize each data modality separately before integration to preserve scale-specific variance.
Apply statistical models (e.g., mediation analysis) to infer regulatory relationships between methylation and expression.
Visualize integrated results using heatmaps with dendrograms or Circos plots for cross-omic interactions.
Assess data missingness patterns in multi-omics datasets and apply imputation cautiously.
Document provenance of each dataset to ensure reproducibility in joint analyses.

Module 9: Reproducibility, Governance, and Data Sharing

Use version-controlled workflows (e.g., Snakemake, Nextflow) to ensure analysis reproducibility.
Archive raw and processed data in public repositories (e.g., GEO, SRA) with MIAME-compliant metadata.
Apply controlled vocabulary (e.g., EDAM, OBI) in metadata to enhance dataset discoverability.
Implement checksums for data files to detect corruption during transfer or storage.
Define data access levels and consent restrictions for human-derived expression data.
Document software versions, parameters, and environment configurations using containerization (e.g., Docker).
Structure project directories following standards (e.g., NIH Data Commons) for team collaboration.
Conduct periodic audit trails of analysis steps to support regulatory or publication review.