Portfolio
NGS pipelines, genomics platforms, and the software behind them.
WES/WGS germline & somatic pipelines
Modular Nextflow DSL2 pipelines for clinical WES and WGS supporting FASTQ, BAM, and VCF entry points, enabling same-day turnaround. Multi-caller variant detection (SNV/Indel via GATK4 and DeepVariant, CNV via CNVkit, SV via Manta) with integrated QC via FastQC, Fastp, and MultiQC.
ACMG/AMP classification engine
Python engine implementing ClinGen/SVI 2023 criteria with HPO/GADO phenotype prioritization, plus automated PDF/HTML clinical reporting per sample.
RNA-seq workflows
STAR/HISAT2 alignment, Salmon/Kallisto quantification, and DESeq2/edgeR differential expression for bulk and targeted transcriptomics.
De novo bacterial/yeast assembly & annotation
End-to-end Nextflow DSL2 pipeline: Kraken2 decontamination, SPAdes assembly, structural/functional annotation (Prokka, BRAKER3, RepeatMasker, tRNAscan-SE, Barrnap, InterProScan, eggNOG-mapper), AMR profiling via RGI/CARD, and SSR marker discovery with MISA.
Species identification correction
Resolved a client-reported species misidentification using fastANI and ortholog-based taxid cross-validation, correcting the final report before delivery.
Sequencing QC dashboard
Nextflow-driven QC pipeline with a live web dashboard tracking sequencing yield against client requirements across multiple runs.
16S rRNA amplicon profiling
ASV-based taxonomic profiling with QIIME2/DADA2, alpha/beta diversity analysis, and differential abundance testing via PERMANOVA and LEfSe.
Shotgun metagenomics
Kraken2/Bracken taxonomic profiling and HUMAnN3 functional pathway analysis, with MetaWRAP for MAG reconstruction and QC.
Polygenic risk score (PRS) platform
Scoring platform covering 43 traits across metabolic, cardiovascular, and complex disease domains — GWAS ingestion, allele harmonization, effect-size standardization, and cross-trait score normalization, backed by a SQLite variant database with a pluggable trait-model scoring engine.
vcfilt — VCF streaming filter
Zero-heap-allocation streaming VCF filter in Go, benchmarked at 147,000 variants/second — 12.2x faster than bcftools 1.18 on an 18GB VCF, with byte-for-byte identical output. MIT-licensed, distributed as a static binary, Docker image, and Singularity container. Published as a bioRxiv preprint, April 2026.
Protein folding molecular dynamics
Long-timescale MD simulations of protein folding dynamics; PCA, clustering, and autoencoders to identify dominant conformational states, validated against DSSP. Contributed to a published study on protein energy landscapes.
Protein-ligand docking
Modeled protein-ligand interactions with AutoDock Vina and RDKit to characterize antibiotic resistance binding pose shifts; GROMACS MD extracting RMSD, RMSF, and binding free energy trajectories.