omics.co.in
💼 Portfolio-Ready Projects End-to-End Scientific Workflows GitHub Blueprint Included

Bioinformatics Capstone Projects & Portfolio Blueprints

Bridge the gap between textbook tutorials and resume-ready computational biology work. Build end-to-end, reproducible genomic projects with public datasets, production-grade Nextflow/Python/R scripts, and professional GitHub repository structures to showcase to biotech employers and academic labs.

CAPSTONE PROJECT 1 • CLINICAL ONCOLOGY

Somatic Driver Mutation Discovery & Drug Match Report in Lung Cancer

BWA-MEM2 GATK Mutect2 Ensembl VEP R maftools

Scientific Objective: Process raw paired-end Whole Exome Sequencing (WES) reads from matched tumor-normal biopsies of lung adenocarcinoma (LUAD) patients, call high-confidence somatic Single Nucleotide Variants (SNVs) and Indels, annotate functional consequences against ClinVar and COSMIC, and generate clinical waterfall and oncoplot visual summaries highlighting therapeutic vulnerabilities (EGFR, ALK, KRAS).

5-Phase Execution Roadmap

  1. Phase 1: Raw Quality Control & Adapter Trimming: Run fastp with --detect_adapter_for_pe, generate MultiQC report, ensure mean base quality > Q30.
  2. Phase 2: Alignment & Preprocessing: Align with bwa-mem2 against GRCh38, sort with samtools sort, mark optical/PCR duplicates using gatk MarkDuplicates, and apply Base Quality Score Recalibration (BQSR).
  3. Phase 3: Somatic Mutation Calling: Execute gatk Mutect2 in paired mode using a Panel of Normals (PoN) and gnomAD germline resource, followed by gatk FilterMutectCalls to discard oxoG artifacts and strand bias.
  4. Phase 4: Variant Annotation & Filtering: Annotate with Ensembl VEP (SIFT, PolyPhen-2, ClinVar), filter for FILTER == 'PASS' and Allele Frequency < 0.001 in population databases.
  5. Phase 5: Downstream Oncogenomics & Visualization: Convert VCF to MAF format using vcf2maf, generate co-mutation matrix and lollipop mutation diagrams in R with maftools, and query OncoKB for FDA-approved targeted therapies.
# Sample Phase 3 Command:
gatk Mutect2 -R reference.fasta -I tumor.bam -I normal.bam -normal NORMAL_SAMPLE_ID \
  --germline-resource af-only-gnomad.vcf.gz --panel-of-normals pon.vcf.gz -O raw_somatic.vcf.gz
gatk FilterMutectCalls -R reference.fasta -V raw_somatic.vcf.gz -O filtered_somatic.vcf.gz
CAPSTONE PROJECT 2 • METAGENOMICS

Human Gut Microbiome Dynamics Under Antibiotic Perturbation

DADA2 QIIME 2 phyloseq ANCOM-BC

Scientific Objective: Quantify the collapse and recovery of microbial diversity in longitudinal stool samples collected before, during, and after broad-spectrum antibiotic administration (Ciprofloxacin). Infer exact Amplicon Sequence Variants (ASVs), assign SILVA taxonomy, calculate alpha/beta diversity, and identify microbial taxa significantly depleted or resilient.

5-Phase Execution Roadmap

  1. Phase 1: Read Inspection & Parametric Error Learning: Plot quality profiles, apply DADA2 filterAndTrim() (truncQ=2, maxEE=c(2,2)), and learn instrument error matrices with learnErrors().
  2. Phase 2: ASV Dereplication & Bimera Removal: Infer exact sample sequences with dada(), merge forward and reverse pairs with mergePairs(), and remove PCR chimeras with removeBimeraDenovo().
  3. Phase 3: SILVA 138 Taxonomic Classification: Assign bacterial kingdom-to-genus taxonomy with assignTaxonomy() against SILVA v138.1 reference database.
  4. Phase 4: Community Diversity Analytics: Import into R phyloseq, compute Alpha diversity (Shannon, Simpson, Chao1) with Wilcoxon rank-sum tests, and perform Principal Coordinates Analysis (PCoA) on Bray-Curtis dissimilarity.
  5. Phase 5: Differential Abundance & Functional Metagenome Imputation: Apply ANCOM-BC to identify differentially abundant genera, and predict functional pathways (e.g., Short-Chain Fatty Acid biosynthesis) using PICRUSt2.
CAPSTONE PROJECT 3 • SINGLE-CELL GENOMICS

Single-Cell Immune Microenvironment Landscape in Immunotherapy Responders

Seurat v5 Harmony SingleR CellChat

Scientific Objective: Dissect cellular heterogeneity in single-cell transcriptomes of tumor-infiltrating lymphocytes (TILs) from melanoma patients receiving anti-PD-1 checkpoint inhibitors. Uncover molecular markers of exhausted CD8+ T cells (PDCD1, HAVCR2, LAG3) and reconstruct intercellular ligand-receptor communication networks.

5-Phase Execution Roadmap

  1. Phase 1: Cell QC & Doublet Scrubbing: Filter cells by feature count (500 < nFeature < 5000), mitochondrial reads < 10%, and remove multiplet droplets using DoubletFinder.
  2. Phase 2: Normalization & Harmony Integration: Apply variance-stabilizing transformation with SCTransform(), regress out cell cycle scoring, and integrate patient samples using Harmony.
  3. Phase 3: Graph Clustering & Automated Annotation: Construct shared nearest neighbor graph, identify clusters via Leiden algorithm, and annotate cell identities with SingleR against Human Primary Cell Atlas.
  4. Phase 4: Pseudotime & T-Cell Exhaustion Dynamics: Subcluster CD8+ T cells and model transcriptional trajectory from effector to terminally exhausted states using Monocle3.
  5. Phase 5: Ligand-Receptor Intercellular Signaling: Infer cell-cell communication networks between tumor-associated macrophages (TAMs) and T cells using CellChat.

🐙 The Ideal Bioinformatics GitHub Repository Blueprint

Structure your capstone repositories according to FAIR computational biology standards to stand out to hiring managers:

my-bioinformatics-project/
├── README.md               # Badges, graphical abstract, motivation, quickstart, citation
├── LICENSE                 # MIT or Apache 2.0 open source license
├── environment.yml         # Reproducible Conda/Mamba dependencies with pinned versions
├── Dockerfile              # Containerized build with all system dependencies
├── workflow/
│   ├── main.nf            # Nextflow DSL2 orchestration script (or Snakefile)
│   └── nextflow.config    # Execution profiles (local, slurm, docker, awsbatch)
├── scripts/
│   ├── 01_qc.sh           # Standalone modular processing scripts
│   ├── 02_align.py
│   └── 03_downstream.R
├── tests/                  # CI/CD GitHub Actions test workflows using toy datasets
└── .gitignore              # Ignores *.bam, *.fq.gz, *.vcf.gz, work/, .nextflow/