Use for any task involving the HDMA (Human Development Multiomic Atlas) repository — using its data (fragments, count matrices, Seurat/ArchR/BPCells objects, ChromBPNet models, contribution scores, bigwigs, motifs, motif instances, caCREs, ABC loops) or its code (preprocessing, global analysis, ChromBPNet training/interpretation, enhancer/ABC analysis, variant scoring, figure reproduction). Triggers on downloading data from Zenodo, finding files for a given cell type or organ, loading an object, reprocessing raw SRA data, loading tracks in a genome browser, running or adapting a pipeline step, finding which script made a figure, or setting up R/conda envs. This is a router skill — read the relevant file under references/ before acting, because the repo has non-obvious conventions (data lives on Zenodo not in-repo; cluster files are named by Cluster_ChromBPNet; scripts read a gitignored ROOT_DIR.txt).
2026-07-17
