Displaying 9 documented projects.
Addresses the subjective interpretation of enrichment results and the disconnect between knowledge bases.
A modular framework that augments gene set enrichment (GO/KEGG ORA) with knowledge graph retrieval (~39M relations from STRING, KEGG, Reactome, PubTator) and parallel literature search from PubMed and the web. An LLM pipeline extracts biomedical relations and synthesizes narrative reports, streamed in real time to a Vue 3 frontend with an interactive D3 network graph and result-grounded chat assistant.
A biomedical-themed portfolio built with Nuxt 3, where the visual narrative follows the story of cancer — from mutation to uncontrolled growth. Features a D3-powered skill graph inspired by Path of Exile, a git-graph experience timeline, terminal-aesthetic UI with DNA backgrounds, and an automated data layer that validates tag dependencies and generates a formatted CV PDF via Playwright in CI.
A Nextflow-orchestrated pipeline that chains Sequenza (CNV calling & purity/ploidy estimation), scarHRD (homologous recombination deficiency scoring), and PyClone-VI (clonal architecture inference) into a single reproducible workflow for WES tumor samples, with automated batch processing and structured output for downstream analysis.
Redesigned the lab website with a terminal-inspired UI. Enhanced member profiles with GitHub-style pixel avatars and contact links, added publication filtering and sorting by year, and integrated an interactive map with detailed location info. Deployed on Nginx with HTTPS via Certbot.
Comprehensive WES-based comparative genomics study of lung adenosquamous carcinoma (ASC) with lung adenocarcinoma and squamous cell carcinoma. Conducted SNV/INDEL, CNV, HRD, and mutational signature analyses, integrated multi-cohort datasets, and explored subtype- and component-level genomic patterns to investigate tumor evolution and cell-of-origin hypotheses.
A reproducible toolkit of shell scripts and Python utilities covering the full NGS workflow: TCGA/SRA data download, BAM-to-FASTQ conversion with QC, WES variant calling (MuSE, MuTect2, VarScan, Pindel) with VCF merging and VEP/MAF annotation, RNA-seq quantification (kallisto, STAR + RSEM), methylation DMP/DMR analysis, and conda environment definitions for each pipeline.
Spatial transcriptomics analysis of 10x Xenium data, exploring both transcript-based and image-based cell segmentation approaches. Downstream analysis with Scanpy covers clustering, spatial visualization, and feature matrix construction from segmented cells.
End-to-end 16S amplicon analysis pipeline built on QIIME2 and DADA2, covering QC, denoising, ASV construction, phylogenetic analysis, and alpha/beta diversity. Integrates BLAST-based NCBI taxonomy refinement for species-level resolution and PICRUSt2 functional prediction. Config-driven and deployed on Linux/GCP.