The world's first multi-omic AI engine for clinical cancer genomics.
GC-Somatic identifies the true, actionable driver mutations in a tumor while filtering out the false positives that slow clinical workflows — reading Illumina, PacBio HiFi, and Oxford Nanopore data, including native DNA methylation, in a single unified engine.
Legacy bioinformatics stacks are fragmented and inaccurate.
10 million cancer deaths occur annually — and somatic variant-calling software hasn't kept pace with the sequencing hardware generating the data.
Unsustainable noise floor
Excessive false-positive variant calls force pathologists into thousands of hours of manual verification every year — adding cost and diagnostic delay to every case.
Infrastructure fragmentation
Clinical labs patch together non-interoperable tools per sequencing vendor. No unified benchmark exists, and reproducibility breaks down at institutional scale.
Multi-omic blind spots
Existing callers are single-modality — structurally blind to native DNA methylation from ONT Dorado. Critical epigenetic cancer drivers go undetected.
One engine, universal multi-omic processing.
GC-Somatic replaces 3–5 fragmented tools with a single, hardware-agnostic, production-ready inference pipeline.
Unlocking the epigenetic layer in cancer diagnostics.
First-class AI input
The only somatic variant caller to ingest native Oxford Nanopore 5mC methylation signals as a first-class AI input channel — not a post-processing filter.
Automates dual reporting
Simultaneous structural-variant and epigenomic output in a single inference pass, eliminating two-pipeline workflows.
Closes a real clinical gap
Methylation disruption is a primary oncology driver. Callers that ignore it miss critical cancer signals at CpG sites — GC-Somatic doesn't.
5mC probability encoded as an AI channel alongside sequence — methylation shapes every call.
MM/ML BAM tags · ONT Dorado · CpG methylation-aware filtering
No competitor has replicated this capability as of June 2026.
Eclipsing the industry standard on the CASTLE benchmark.
Independent head-to-head evaluation against DeepSomatic, the cancer-genomics AI published by Google in Nature Biotechnology.
6 matched tumor/normal CASTLE cell lines · Illumina + ONT + HiFi · bcftools isec · evaluated 2026-06-24, corrected baseline 2026-07-12
11 AI model domains — 6 validated, 5 robustness-tested.
Empirically validated on real matched data for 6 of 11 sequencing / library-prep configurations. The remaining 5 demonstrate architecture-level robustness on zero-padded Illumina data, ahead of real-sample validation.
No DeepSomatic equivalent exists for 7 of these 11 domains. Cross-validated across six cancer types: lung adenocarcinoma, breast adenocarcinoma, BRCA1-mutant breast, HER2+ breast, triple-negative breast, and copy-number-instability breast cancer.
Deep technical and structural defensibility.
| Capability | GC-Somatic | DeepSomatic (Google) | Mutect2 (GATK) | Strelka2 |
|---|---|---|---|---|
| Illumina WGS / WES | ✓ | ✓ | ✓ | ✓ |
| PacBio HiFi | ✓ + Phasing AI | ✓ | ✗ | ✗ |
| ONT long-read | ✓ Proprietary | ✓ | ✗ | ✗ |
| Native 5mC methylation | ✓ World first | ✗ | ✗ | ✗ |
| cfDNA liquid biopsy | ◐ OOD F1=0.9626 | ✗ | ✗ | ✗ |
| FFPE artifact suppression | ◐ OOD F1=0.9620 | ✓ limited | ✓ limited | ✓ limited |
| CH / CHIP correction | ✓ AI channel | ✗ | ✗ | ✗ |
| Commercial license | ✓ Enterprise | Open source | Open source | Open source |
18–24 months estimated to replicate the architecture. Patent filings in progress. CLIA revalidation cost ($50–200K) creates durable switching costs once a lab adopts GC-Somatic.
$43B+ total addressable market.
Four converging market segments — all addressable today with GC-Somatic.