Research package · bioRxiv preprint
MINGL
Implemented a scverse-compatible research package and contributed as a named bioRxiv co-author.
Overview
MINGL is a research package for probabilistic cell-type classification in multiplexed tissue imaging. My contribution centered on implementing package tooling that fit into a real lab workflow rather than live only in notebooks.
My contribution
Implemented the scverse-compatible package: GMM classification, 13 tool functions, 13 plotting functions, and lab workflow integration.
Problem
Cell-type classification in multiplexed spatial proteomics often depends on manual gating or brittle heuristics, which makes uncertainty harder to represent and slows down research iteration.
Approach
- Implemented Gaussian Mixture Model-based classification for probabilistic annotation of cells in multiplexed imaging data.
- Published the package in a scverse-compatible format so it could slot into existing bioinformatics workflows.
- Added 13 tool functions and 13 plotting functions spanning gradient, border, and heterogeneity analyses.
Result
Implemented and shipped MINGL as a usable research package, and contributed to the bioRxiv preprint as a named co-author.