Built

Projects

01 / bio

RiboWin

Modular pipeline for computational discovery and translational validation of small ORFs (sORFs) and microproteins from multi-omics Ribo-seq and RNA-seq data.

Built a five-phase pipeline (sORF discovery → RNA-seq alignment → Ribo-seq processing → translation evidence → translation efficiency) with YAML-based single-command execution.
Eliminated hardcoded paths and implemented empirical P-site offset calibration per sample, with a mandatory two-replicate concordance filter.
Benchmarked on human cell lines (HeLa, keratinocyte, fibroblast) aligned to GRCh38; resolved a failure affecting 16,000+ ORFs to improve reliability.
Fully reproducible: auto-downloads FASTQs from SRA, idempotent phases, precomputed test data hosted on Zenodo.

02 / bio · ml

GitHub →

LUAD Survival & Biomarker Study using ML

Multi-omics survival prediction for Lung Adenocarcinoma, integrating RNA-seq, mutation, and clinical data from TCGA.

Trained and evaluated Random Forest, SVM, and Voting Classifier models (~62-66% accuracy) on integrated multi-omics datasets.
Ranked genes by feature importance and identified survival-associated biomarkers including BMP5, ZNF555, GGTLC1, and CTLA4.
Ran Kaplan-Meier survival stratification and g:Profiler functional enrichment, linking key genes to apoptosis and immune-response pathways.
Deployed an interactive Streamlit dashboard for exploring results (confusion matrices, ROC curves, clustered heatmaps).

03 / bio · srfp

GitHub →

PRMT5 and Retrotransposon Regulation

IASc SRFP 2025 project investigating how PRMT5 inhibition affects transposable element expression and the interferon response.

Combined pathway-level GSEA (fgsea, DESeq2) with TE-focused analysis (TEtranscripts, Telescope) on GEO dataset GSE236930.
Found strong upregulation of ERV1-family elements (HERV-int most upregulated) with selective, context-dependent LINE-1 activation.
Classified TE loci as intragenic vs. intergenic using BEDTools to flag potential regulatory interactions.
Compared two PRMT5 inhibitor treatments (GSK591, LLY283), finding distinct TE activation patterns between them.

04 / bio

GitHub →

ChIP-Sequencing (H3K27ac in K562 cells)

Full ChIP-seq analysis pipeline identifying and annotating H3K27ac enhancer/promoter marks in K562 leukemia cells.

Processed ENCODE ChIP-seq data (H3K27ac BAM + input control) with HOMER, BEDTools, and samtools on Ubuntu.
Annotated 43,034 peaks to ~22,000 genes; filtered to 5,603 high-score genes and a top-1,000 gene set for enrichment analysis.
Ran g:Profiler functional enrichment, surfacing transcription regulation, chromatin organization, and cell-cycle pathways.
Visualized H3K27ac signal near key leukemia genes (MYC, GATA1, TP53) using IGV.

05 / mkts · ongoing

Trading Platform

Personal Python-based algorithmic trading system for Indian markets.

Currently in development — Building a proprietary algorithmic trading system for the Indian markets — starting with NSE equities, extending to currency derivatives and options. The pipeline covers feature engineering (VWAP deviation, opening-range breakout, ADX regime detection), triple-barrier labeling for realistic win/loss/timeout classification, regime-conditional signal generation, and an event-based backtester modeling real Indian transaction costs (brokerage, STT, slippage) with hard risk controls (daily loss limits, consecutive-loss circuit breakers). Data and execution are integrated via the Groww Trading API, with a Streamlit dashboard for live strategy monitoring. Next: walk-forward validation on real market data and an XGBoost layer for signal confidence scoring.