SvABA Viewers

Dark, compact table-first viewer for bps.txt.gz files. Sortable columns, numeric filters (somlod, maxlod, qual, span range), genomic region search, chip filters for type/evidence/conf/somatic, summary histograms, per-sample detail panel. Handles both legacy 41-col and SvABA2.0 51-col formats. Drag-and-drop or file picker.

Browser-side SQL explorer for the SQLite r2c database emitted by svaba run --dump-reads + svaba postprocess (v4 and later). Load the .db file and run arbitrary SQL queries client-side via sql.js — no server, no infrastructure. Sidebar has canned queries for the common asks (top contigs, PASS subsets, reads supporting a specific bp_id, support breakdowns) plus a schema reference. Press Cmd/Ctrl+Enter in the editor to run. Results render as a sortable HTML table (capped at 5k rows for DOM sanity — refine with LIMIT for tighter views).

Browser-side re-plotter for the older structured r2c TSV emitted by pre-v4 svaba run --dump-reads. New runs produce r2c.db instead — use the DB Explorer above. This page still works on historical .r2c.txt.gz files (and on the legacy .r2c.pass.txt.gz / .r2c.pass.somatic.txt.gz filtered subsets).

Original ASCII-plot viewer for *.alignments.txt.gz from svaba runs before the r2c migration. New runs no longer produce alignments.txt.gz — prefer the R2C Explorer above. Kept for rendering historical outputs.

Interactive viewer for BAM learning data — per-read-group insert-size distributions emitted by svaba run as ${ID}.*.learn.tsv.gz. Overlaid + faceted histograms, sortable summary table (n, mean, median, SD, percentiles), RG chip filters, click-to-solo. Drag-and-drop the gzipped TSV to explore.

Explore per-region timing from ${ID}.runtime.txt. Sort regions by runtime, contigs, or breakpoint count; filter by chromosome, region, runtime threshold, or contig / discordant-read minimums. Prominent runtime-distribution histogram plus auxiliary contig / discordant histograms with independent log10 toggles. Click a region's locus to jump IGV to it (requires IGV on port 60151). Standard-chr filter drops decoy / alt / random / HLA contigs.

Side-by-side comparison of two svaba runs (e.g. before/after a parameter change, or tumor vs matched). Load two bps.txt.gz files, match calls by locus, and see which are shared vs unique to each run. Charts for agreement rates and score distributions.

Score svaba against the sim/ simulator ground truth. Load truth.bedpe + a bps.txt.gz; sweep a score threshold (somlod / maxlod / qual) to get Precision-Recall and Recall-vs-FalsePositives (FROC) curves, an operating-point slider, and recall broken down by SV type and size bin. Click TP/FP/FN for IGV- linked variant lists; optional blacklist BED masks regions. Streams large gzip, auto-restricts to the truth's chromosomes.

Track how svaba accuracy and compute change across git commits. Point it at a sim panel folder (or all of /Volumes/wala24T/sim); it reads every svaba_runs_<hash>/{run_meta.json,benchmark_summary.tsv} and plots recall/precision/F1 and wall/CPU vs coverage, one line per commit, filterable by panel & purity, plus a per-commit compute bar and an accuracy-vs-compute trade-off scatter. Per-commit somlod / maxlod / contig_conf sliders re-score recall/precision live (no re-run) so you can tune and compare cutoffs per svaba version, plus an evidence-type (DSCRD/ASSMB/…) call filter. Also a per-event matrix (TP/FN/FP) showing which commits got each event right/wrong, with IGV links — loads the per-BAM *.events.tsv.

Front-end for running svaba on GCP scatter-gather (scripts/svaba_cloud.sh). Fill a form (project, disk, BAMs, machine type, partitions) and get the exact commands — precompute --bam-params, launch, monitor, gather, teardown — plus a live cost estimate (VM picker, on-demand vs Spot, memory sanity) and a monitor dashboard (paste the done-marker listing → progress). Client-side only.

The original light-theme BPS viewer. Filter by LOD, confidence, event type, and evidence. Per-sample support drill-down. Requires app.js and styles.css alongside it.