Bioinformatics Analyst / Computational Biologist · MS Bioinformatics

Open to Bioinformatics Analyst, Computational Biologist, and Bioinformatics Scientist roles — especially at academic cancer centers.

📧 sathvikappagana@gmail.com · LinkedIn

I build reproducible, defensible bioinformatics pipelines — each validated against known biology as a positive control, honest about where methods fail instead of smoothing it over, and treating a well-measured null result as a real finding.

My core work is one connected story, deliberately built so each project answers what the previous one couldn't:

• Bulk RNA-seq found the ER+/TNBC subtype signature by averaging across all cells.
• Single-cell localized that signature to the malignant epithelial compartment and proved malignancy cell-by-cell via copy-number inference.
• Spatial deconvolution put the compartments back into tissue, showing the tumor–immune architecture itself differs by subtype — TNBC infiltrated, ER+ excluded.

• rnaseq-brca-pipeline — Bulk RNA-seq of BRCA subtypes (STAR → featureCounts → DESeq2 → clusterProfiler); recovers canonical ER+/TNBC signatures in the correct direction.
• scrna-tme-pipeline — scRNA-seq of the tumor microenvironment (31,265 cells; Scanpy/Harmony); inferCNV confirms malignancy cell-by-cell.
• spatial-tme-pipeline — Reference-based spatial deconvolution (RCTD/SPOTlight) across 6 Visium sections, validated against pathologist annotation.

• multiomics-ml-brca — Permutation-validated multi-omics ML for relapse (RNA-seq + DNA methylation, TCGA-BRCA); a 1,000× leak-free noise floor shows methylation adds no gain beyond expression — the measurement is the deliverable, not the direction.
• spatial-trimm-project — Spatial cell-motif discovery (TrimNN) with a permutation null correcting abundance bias, plus fixes for cell-type confounding and pseudoreplication.
• atac-seq-pipeline — Differential chromatin accessibility (bowtie2 → MACS2 → DESeq2 → HOMER); independently recovers the FOXA1 pioneer factor as a positive control.
• gatk-somatic-variant-calling — GATK4 Best-Practices tumor–normal WES (hg38; bwa-mem2 → Mutect2 → Funcotator → maftools) on a real open-access patient case.

Every repo ships a pinned conda environment, a data-download script, and a README that regenerates the figures — clone and run.

Languages — Python · R · Bash Single-cell & spatial — Scanpy · Harmony · inferCNV · RCTD · SPOTlight · squidpy Bulk & epigenome — STAR · featureCounts · DESeq2 · bowtie2 · MACS2 · HOMER · clusterProfiler Variants — GATK4 · bwa-mem2 · Mutect2 · maftools ML & statistics — scikit-learn · permutation testing · elastic-net / random forest · nested cross-validation Pipelines & infra — Snakemake · SLURM · AWS EC2