molamola

Outputs

Each run writes a single self-contained HTML artefact into --out (default: the parent directory of --vcf). The filename and contents differ by the auto-detected mode:

• SV mode → <sample>.report.html (circos plot + linear cytoband SV map embedded as base64 PNG data URIs, plus a collapsible run-metadata block).
• Compound-het mode → <sample>.compound_het.report.html (one phased-haplotype panel per gene, embedded the same way; auto-select runs split into a strict and extended section).

No separate PNG files are written; figures live inside the HTML. Right-click → save to extract a PNG locally if you need one — each compound-het panel’s PNG carries the gene symbol baked in (top-left, bold) so a saved image is self-describing.

Reference data

All bundled in molamola/data/:

• cytoBand.txt.gz (hg38) and cytoBand.t2t.txt.gz (T2T-CHM13v2.0) — UCSC cytoband annotations for SV mode.
• canonical_exons.hg38.tsv.gz — MANE Select v1.5 canonical-transcript span + per-exon coordinates (~19,200 protein-coding genes).
• clinvar.hg38.tsv.xz — molamola’s reduced ClinVar TSV (~13 MB; xz-compressed; release date logged in each report’s run-metadata). The --clinvar flag accepts either this TSV or NCBI’s raw ClinVar VCF (auto-detected by extension).

molamola is bundled-only by design — no auto-download, no online lookups. All bundled refs are reproducibly regeneratable from public sources by running scripts/derive_canonical_exons.py and scripts/derive_clinvar_for_molamola.py.

Circos plot (SV mode)

Outer ring: cytoband ideogram (default pyCirclize colours). Inner ribbons: each PASS BND, line-thickness scaled by SUPPORT, colour by VAF (plasma, saturating above 0.7). Noise-flagged BNDs render grey/dashed at low alpha so the eye goes to the candidate signal.

Genome SV map (SV mode)

One row per chromosome, top-down chr1 → chrY:

• Chromosome bar (cytoband ideogram, greyscale: white → very-dark-grey, centromere = black).
• Four density strips above the bar, one per non-BND SV type (INS = blue, DEL = red, DUP = green, INV = purple). Each strip is 1 Mb-binned; alpha encodes count (sqrt-scaled, single-event bins still visible).
• BND arcs above the strips, with apex height proportional to row gap and span. Same VAF / SUPPORT encoding as the circos ribbons.

Phased-haplotype panel (compound-het mode)

One panel per plotted gene. Top-down:

• Title — gene symbol baked into the PNG (top-left, bold) so saved images are self-describing.
• Canonical-transcript exon track — IGV-style blue rectangles connected by a thin grey line; coordinates from the bundled MANE Select TSV.
• Mint phase block — one rectangle per WhatsHap PS group, spanning both haplotypes; off-edge arrows when the block stretches past the gene window.
• H1 / H2 hap lines — grey horizontal rules; H1 above H2.
• Missense lollipops — one filled circle per phased het missense canonical-transcript variant, hanging downward from its hap line. Colour by ClinVar bucket (P/LP #c0143c, VUS #f4a013, conflicting #d9c200, benign #5fa860, no-ClinVar grey).
• Synonymous-variant ticks — x markers on the hap line for canonical-transcript variants that aren’t missense, included as context.
• Chip row above the figure — missense / total / blocks / trans / cis stats.
• Legend — at the bottom; ClinVar buckets + the synonymous tick.

HTML report

A single self-contained HTML file (no server, no external CSS/JS, opens offline). Figures are embedded as base64 PNG data URIs. A collapsible run-metadata block at the bottom shows mode-specific provenance:

• SV mode: caller, filter mode, baseline coverage, cov-anomaly threshold, BND noise breakdown.
• Compound-het mode: ClinVar source + release date, canonical-exons source, VCF parse refusal counts, selection rule, plotted-vs-scanned gene counts.

Designed as a portable review surface — emailable, archivable, no install needed for the reader.

This site is open source. Improve this page.