Track Record
Problems We've Solved
Our work focuses on transforming manual or fragmented scientific workflows into reproducible, scalable systems.
MRMpipe
Flagship ProductAutomated peak detection, overlapping signal separation, and ML-based quality scoring for targeted LC-MS workflows.
Problem
Analysts spend hours manually reviewing and adjusting peak boundaries. Results vary between operators and across batches.
What We Built
End-to-end pipeline with BEMG deconvolution, DTW cross-sample consistency, and ML confidence scoring.
Outcome
Reduces manual chromatogram review and delivers consistent quantification across batches.
SysQuan
Scientific DatabaseSearchable peptide panel database with QC review and automated export for MRM/PRM proteomics.
Problem
Thousands of peptide panels tracked across scattered spreadsheets with no centralized search.
What We Built
Custom database with search, QC review, and automated transition list export.
Outcome
Single source of truth for peptide panels. Lookup time from minutes to instant.
MHCvalidator
ML Rescoring EngineMachine learning rescoring for immunopeptidomics peptide identification.
Problem
Low-confidence peptide IDs couldn't be separated from false positives using conventional scoring.
What We Built
ML model trained on domain-specific features rescores every peptide-spectrum match.
Outcome
Significantly more true peptides recovered at the same FDR. Published in Nature Communications.
Other Work
MhcVizPipe
Interactive QC Dashboard
No way to quickly assess immunopeptidomics QC across experiments. Built an interactive dashboard for allele coverage, contamination, and reproducibility at a glance.
RHybridFinder
R Package • CRAN
Detecting hybrid HLA ligands required manual proteome cross-referencing. Automated the scan into an R package available on CRAN.
Full publication list on PubMed.