Description

This curriculum spans the full bioinformatics workflow from experimental design to translational handoff, comparable in scope to a multi-phase research program integrating data generation, computational analysis, and cross-functional collaboration across wet and dry labs.

Module 1: Defining Biological Objectives and Study Design

Select appropriate tissue types and developmental stages for sampling based on the biological hypothesis, balancing relevance with availability and ethical constraints.
Determine case-control or time-series experimental design depending on whether the goal is differential expression or dynamic gene behavior analysis.
Establish sample size requirements using power calculations informed by expected effect sizes and variability from pilot data or literature.
Decide between bulk RNA-seq and single-cell RNA-seq based on cellular heterogeneity concerns and resolution needs.
Negotiate access to clinical metadata while complying with patient privacy regulations and institutional review board requirements.
Coordinate with wet-lab teams to standardize collection, preservation, and shipping protocols across multiple sites.
Define primary and secondary endpoints for clustering outcomes, such as biomarker identification or pathway enrichment.
Document experimental design decisions in a data management plan to ensure reproducibility and audit readiness.

Module 2: Data Acquisition and Quality Control

Validate raw sequencing data integrity using FastQC and verify expected read lengths, GC content, and adapter contamination levels.
Implement automated pipelines to flag samples with low sequencing depth or high ribosomal RNA content for exclusion.
Compare alignment rates across samples using STAR or HISAT2 to detect batch effects or technical outliers.
Apply sample-level filtering thresholds based on gene detection rates and total read counts to remove low-quality libraries.
Integrate external datasets only after confirming compatibility in library preparation, sequencing platform, and read length.
Document all quality control decisions and thresholds in a standardized report for audit and replication.
Use principal component analysis on raw counts to identify unintended sources of variation such as sex or batch.
Establish a data freeze point after QC to prevent uncontrolled reprocessing during downstream analysis.

Module 3: Preprocessing and Normalization Strategies

Choose between TPM, FPKM, and DESeq2's median-of-ratios depending on downstream use in clustering versus differential expression.
Apply log-transformation with pseudocount adjustment, evaluating the impact on distribution symmetry and outlier sensitivity.
Remove lowly expressed genes using mean-variance trends and minimum counts-per-million thresholds.
Correct for library size variation using TMM normalization when combining datasets from different sources.
Assess the influence of housekeeping genes on normalization stability across conditions.
Decide whether to preserve or remove batch effects at this stage based on study objectives and confounding risks.
Implement gene filtering based on variance across samples to reduce noise before clustering.
Validate normalization efficacy using density plots and MA plots to confirm centering and spread consistency.

Module 4: Dimensionality Reduction and Feature Selection

Compare PCA, t-SNE, and UMAP embeddings to assess preservation of global versus local structure in gene expression space.
Select the number of principal components using elbow plots or cumulative variance thresholds, typically 80–90%.
Evaluate feature stability in high-variance gene lists across resampling iterations to prevent overfitting.
Apply surrogate variable analysis (SVA) to capture hidden confounders not accounted for in experimental design.
Use mutual information or correlation-based filtering to remove redundant genes before clustering.
Validate dimensionality reduction outputs by overlaying known cell type markers or experimental conditions.
Balance computational efficiency and interpretability when choosing between linear and nonlinear methods.
Store reduced feature sets with metadata linking back to original gene identifiers and selection criteria.

Module 5: Clustering Algorithm Selection and Execution

Choose between k-means, hierarchical, and DBSCAN clustering based on expected cluster shapes and density distribution.
Determine optimal k using the gap statistic, silhouette score, or elbow method, validating across multiple metrics.
Run consensus clustering with bootstrapping to assess cluster stability and reduce algorithmic sensitivity.
Adjust distance metrics (Euclidean, Pearson, Spearman) based on data distribution and biological interpretability.
Handle outliers by either isolating them or using robust clustering methods that accommodate noise.
Parallelize clustering jobs across gene subsets or parameter grids to manage computational load.
Preserve intermediate clustering results for parameter tuning and comparative analysis.
Implement reproducible execution using containerized environments with fixed random seeds.

Module 6: Cluster Validation and Biological Interpretation

Calculate cluster cohesion and separation using within-cluster sum of squares and between-cluster distances.
Compare clustering solutions using adjusted Rand index when integrating multiple algorithms or parameters.
Annotate clusters with overrepresented GO terms using hypergeometric tests and correct for multiple testing.
Map clusters to known pathways using KEGG or Reactome and evaluate enrichment significance with FDR thresholds.
Identify hub genes within clusters using intramodular connectivity measures from WGCNA.
Validate cluster robustness by subsampling genes and recalculating membership consistency.
Integrate transcription factor binding data to infer regulatory drivers of cluster-specific expression.
Flag clusters with ambiguous or mixed biological signatures for re-evaluation or splitting.

Module 7: Cross-Dataset Integration and Reproducibility

Apply ComBat or Harmony to correct batch effects while preserving biological variation across datasets.
Use reciprocal PCA or CCA to align gene expression spaces from different studies or platforms.
Validate cluster conservation by projecting new data onto existing clustering models using label transfer.
Assess generalizability of clusters by testing enrichment in independent cohorts with similar phenotypes.
Document version control for gene annotations and reference databases to ensure consistency.
Standardize gene identifiers across datasets using biomaRt or UniProt mapping with conflict resolution rules.
Archive processed data matrices and clustering labels in standardized formats (e.g., HDF5, Seurat objects).
Implement checksums and metadata logs to track data transformations across integration steps.

Module 8: Reporting and Translational Handoff

Generate cluster-specific gene lists with fold changes, p-values, and functional annotations for wet-lab validation.
Produce publication-ready heatmaps with dendrograms, sample annotations, and color-coded metadata.
Export interactive visualizations using Plotly or Shiny for stakeholder exploration.
Define minimal reporting standards including clustering parameters, software versions, and preprocessing steps.
Prepare data packages for deposition in GEO or ArrayExpress with MIAME-compliant metadata.
Coordinate with biologists to prioritize candidate genes for knockout or overexpression assays.
Document limitations such as potential overclustering or underrepresentation of rare cell types.
Establish a change log for all analytical updates to support audit and regulatory review.