#22 · ML / AI methods

Recording analysis provenance and decisions in computational biology workflows

The common thread across these preprints is making computational analysis in biology traceable. EMMA automatically captures annotation releases, software versions, and parameters during functional enrichment analysis, while ProteoScopeR records quantitative proteomics decisions and compares normalization, missing-data, and differential workflows. Murmurent adds traceability and multi-user project structure to agentic AI infrastructure for biomedical labs, and ScProteoAgent retains analysis state and matrix history while translating natural-language requests into single-cell proteomics calculations. ATACofthesnake and the CellProfiler microglial morphology pipeline emphasize scalable, reproducible workflow design for chromatin accessibility and clustering analyses rather than provenance per se. What is new here is the shift of provenance and decision logging into a built-in component of the analysis tool itself, although the items describe workflow design and do not evaluate whether recorded provenance changes downstream conclusions.