mSigSDK: browser-native computation of mutational signatures

Mutational-signature analysis compares a sample’s somatic mutation pattern against reference signatures to estimate exposures. Mature R and Python packages exist, but each requires local installation and its own input format, and the choice of fitting tool meaningfully affects results. Triangulating across tools is therefore desirable but operationally expensive, and no standard machine-readable format captures an analysis’s parameters, review evidence, and provenance. mSigSDK is a JavaScript SDK that loads in tested desktop browsers via a single dynamic import, with no installation. Analysis runs locally: spectra and MAF rows stay on-device, though optional reference-context helpers may transmit genomic coordinates to public endpoints unless strict-local mode is enabled. The SDK orchestrates four established tools (SigProfilerAssignment, MuSiCal, deconstructSigs, sigminer) via a uniform adapter layer, adds browser-native NNLS fitting and NMF-based exploratory extraction, emits configurable rule-based review evidence, and serializes portable, schema-validated reports with built-in visualization and export. A zero-install demo reached a rendered local report in ~1.82 s. Each adapter reproduced its package's local execution on 38 PCAWG Lung-AdenoCA SBS96 spectra — exactly for three tools, to floating-point precision for the fourth. Across the tested fixtures, the converter matched SigProfilerMatrixGenerator 1.3.6 exactly for SBS96, SBS1536, and DBS78 and for ID83 class assignment when SigProfilerMatrixGenerator-derived repeat/microhomology annotations were supplied to mSigSDK. Across 2,700 synthetic spectra at three noise levels, mean exposure cosine exceeded 0.96 for all tools, but mean per-spectrum active-signature F1 ranged from 0.532–0.924 at 10% noise. On PCAWG, all tools reconstructed spectra well (mean cosine 0.982–0.994) while calling 4.13–17.39 active signatures per sample. Warm-start refitting took 18.2–73.7 ms (Windows) and 22.2–44.6 ms (macOS) for 120 samples, and 1.80–6.81 s (Windows) and 2.12–5.89 s (macOS) for 300 samples/40 signatures. mSigSDK brings multi-tool orchestration, exploratory extraction, rule-based review, and schema-based reporting and visualization to the browser without local installation.

Authors

Institutions

Publication Details

Journal
BMC Bioinformatics
Published
2026-09-30
DOI
https://doi.org/10.1186/s12859-026-06681-z
Primary Topic
Cancer Genomics and Diagnostics
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
article

mSigSDK: browser-native computation of mutational signatures

Aaron Ge, Jonas S. Almeida, Yasmmin Martins, Kailing Chen et al.
BMC Bioinformatics
Cancer Genomics and Diagnostics
article

mSigSDK: browser-native computation of mutational signatures

Aaron Ge, Jonas S. Almeida, Yasmmin Martins, Kailing Chen, Jeya Balasubramanian, Tongwu Zhang, Maria Teresa Landi, Brian Park
article en

Abstract

Mutational-signature analysis compares a sample’s somatic mutation pattern against reference signatures to estimate exposures. Mature R and Python packages exist, but each requires local installation and its own input format, and the choice of fitting tool meaningfully affects results. Triangulating across tools is therefore desirable but operationally expensive, and no standard machine-readable format captures an analysis’s parameters, review evidence, and provenance. mSigSDK is a JavaScript SDK that loads in tested desktop browsers via a single dynamic import, with no installation. Analysis runs locally: spectra and MAF rows stay on-device, though optional reference-context helpers may transmit genomic coordinates to public endpoints unless strict-local mode is enabled. The SDK orchestrates four established tools (SigProfilerAssignment, MuSiCal, deconstructSigs, sigminer) via a uniform adapter layer, adds browser-native NNLS fitting and NMF-based exploratory extraction, emits configurable rule-based review evidence, and serializes portable, schema-validated reports with built-in visualization and export. A zero-install demo reached a rendered local report in ~1.82 s. Each adapter reproduced its package's local execution on 38 PCAWG Lung-AdenoCA SBS96 spectra — exactly for three tools, to floating-point precision for the fourth. Across the tested fixtures, the converter matched SigProfilerMatrixGenerator 1.3.6 exactly for SBS96, SBS1536, and DBS78 and for ID83 class assignment when SigProfilerMatrixGenerator-derived repeat/microhomology annotations were supplied to mSigSDK. Across 2,700 synthetic spectra at three noise levels, mean exposure cosine exceeded 0.96 for all tools, but mean per-spectrum active-signature F1 ranged from 0.532–0.924 at 10% noise. On PCAWG, all tools reconstructed spectra well (mean cosine 0.982–0.994) while calling 4.13–17.39 active signatures per sample. Warm-start refitting took 18.2–73.7 ms (Windows) and 22.2–44.6 ms (macOS) for 120 samples, and 1.80–6.81 s (Windows) and 2.12–5.89 s (macOS) for 300 samples/40 signatures. mSigSDK brings multi-tool orchestration, exploratory extraction, rule-based review, and schema-based reporting and visualization to the browser without local installation.

BMC Bioinformatics
National Institutes of Health (US), Laboratório Nacional de Computação Científica (BR), Division of Cancer Epidemiology and Genetics
Openalex Percentile: Top 16%
Cancer Genomics and Diagnostics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.