Open-source models, all the way down.
SciRouter routes every step of the analysis through the same biomedical model stack already in production for our drug-discovery customers. No proprietary black box — every variant call, every neoantigen rank, every drug recommendation is reproducible. Below, the route table the oncology product runs against.
Why this matters
FidoCure, Vidium, and ImpriMed each ship a slice of this stack as a closed product. SciRouter Oncology treats genomics as data we ingest, not a product we sell exclusively. That means a vet using ImpriMed's chemosensitivity assay alongside a Vidium 120-gene panel — both feeding a single canonical patient record, with a single drug-mutation matrix in the vet view.
Personalized mRNA neoantigen vaccines (SciRouter mRx) follow the same pattern: WES + RNA-seq → DLA typing → neoantigen ranking via pVACtools → AlphaFold-3 confirmation of peptide-MHC binding → mRNA construct → LNP formulation via a manufacturing partner (Vernal Biosciences in Vermont, or UNSW RNA Institute in Sydney for the Rosie protocol). Administered under USDA-CVB conditional licensure or university IACUC trial wrappers.
The two-tier Gemma 4 deployment maps cleanly to the product: Gemma 4 4B always-on for the conversational owner-side explainer; Gemma 4 26B MoE serverless on RunPod for vet-side decision synthesis, document parsing, and the tumor-board agent committee.
For deeper technical reading on the SciRouter platform, see the SciRouter homepage and engineering blog.