bio-core

Organism-agnostic bioinformatics mechanism — the shared foundation for MethylAsk (human methylation), GeneAsk (human variants), and the seagrass plant-epigenomics work.

What lives here (and what doesn't)

bio-core is pure mechanism: file I/O, the context-aware methylation model, coordinate handling, provider/cache interfaces, evidence tiering, and report rendering. It contains no organism-specific knowledge — no clinical databases, no epigenetic clocks, no plant pathways. Those live in the downstream repos that depend on bio-core.

The split is knowledge vs mechanism: a modBAM from a human and from eelgrass are the same file read by the same code (mechanism, here); what a marker means is organism-specific (knowledge, downstream).

Modules

• biocore.methylation.model — context-aware model (cytosine + CG/CHG/CHH context + level) and the weighted-methylation estimator
• biocore.io.bedmethyl — modkit 18-column bedMethyl reader (streaming)
• biocore.io.modbam — ONT modBAM reader: piles up MM/ML methylation calls into MethylSite rows, with CG/CHG/CHH context from a reference FASTA (needs pysam)
• biocore.io.vcf_ops — bcftools wrappers (reheader / merge / biallelic filter)
• biocore.io.fetch — resumable, retrying downloader for slow academic mirrors
• biocore.providers — provider interface, cache, error-tolerant status, evidence tiering
• biocore.report — one report model → HTML/PDF

Provenance

The methylation reader and VCF ops were recovered and refactored from the eelgrass analysis pipeline; the provider/tiering/report layer was lifted from MethylAsk. See docs/PROVENANCE.md.