Think of it like a weather forecast — for mental health risk
A forecast doesn't claim it will definitely rain. It reads the signals in the atmosphere and reports a probability you can act on. NeuroPenetrance applies the same logic to biology.
Instead of a yes-or-no genetic diagnosis, NeuroPenetrance quantifies biological risk using molecular data — giving clinicians and researchers an evidence-based signal they can actually act on.
Biological data
Gene and protein expression from patient-derived iPSC neurons — RNA-seq, proteomics, ATAC-seq.
ML pipeline
Classifiers trained across thousands of molecular features find the patterns that separate vulnerability from resilience.
Risk signal
A ranked, interpretable Penetrance Risk Profile — not a black box — for clinical or pharma use.
Why cellular data, not just a DNA sequence
A separate category of company offers DNA-based polygenic risk scores for psychiatric conditions — Orchid Health and Genomic Prediction for prospective parents, Myriad Genetics' RiskScore for breast cancer risk. These companies work from static, inherited DNA sequence alone.
NeuroPenetrance's approach is fundamentally different in kind: it measures the functional, cellular consequence of a variant in patient-derived neurons — which can reveal how modifier genes, environment-like perturbations, and individual genetic background shape whether a risk variant actually produces disease. That's precisely the biological question a DNA sequence alone cannot answer. It's a genuine scientific differentiator, not just a marketing angle.
The bridge from blood to brain
The Engine 2 proof of concept — described in full on the Evidence page — was deliberately built on a hard test case: cross-sectional whole-blood transcriptomics, a technically noisier, less biologically direct signal than cell-type-specific data from genetically defined cell lines.
The fact that this pipeline already extracts validated, interpretable signal under those harder conditions is strong evidence for its performance once applied to Engine 1's richer, purpose-generated iPSC-derived datasets — neuronal transcriptomic and proteomic data tied to known genetic variants and controlled experimental perturbations, rather than a noisy cross-sectional blood sample.
The compounding data flywheel
The two engines reinforce each other through a data flywheel: every Engine 1 service contract generates proprietary training data that improves Engine 2's predictive power, which in turn increases the commercial value of each new service contract. This compounding advantage is difficult for a generic iPSC contract lab or an academic core facility to replicate — they have the wet-lab capability, but not the AI layer built specifically for this question.
See how the two engines work together → See the validated results →