My venture
NeuroPenetrance: reading the immune biology of depression from a blood sample
NeuroPenetrance measures the activity (RNA) of six immune genes in whole blood and turns it into an interpretable profile for every patient, to help find which patients carry the biology a treatment targets.
The problem
Depression is not one disease, but it is still treated as one
Depression is diagnosed and treated almost entirely by interview and trial and error; there is no blood test. Yet about one patient in four shows low-grade inflammation in the blood, and early studies suggest this group responds differently to treatment. When a drug trial cannot find these patients, a medicine that helps some can look useless for all.
The same idea runs through genetics: carrying a risk is not the same as becoming ill, a gap called incomplete penetrance, which gives the company its name.
The approach
A blood test first, a laboratory engine next
Engine 2: blood immune signature (available now)
Six immune genes measured in whole-blood RNA: three inflammation genes that rise in depression and three lymphocyte genes that fall. The score is simply one group minus the other, with no black box. The PenetranSci app turns it into a calibrated probability with a 95% range, a PDF report and batch analysis for whole studies. It works on samples clinics and trials already store.
Engine 1: patient-derived neurons (in development)
Why does the same genetic risk lead to illness in one person and resilience in another? Engine 1 grows a person's own cells into neurons from stem cells and measures what their biology does, building on my Fulbright work at the Max Delbrück Center in Berlin.
The evidence
Tested on studies it had never seen, and reported honestly
8 public whole-blood RNA studies of depression and healthy controls, from different labs and technologies.
The six-gene signature was built on one study, then tested unchanged on 7 others (0.50 = chance, 1.00 = perfect). Modest, about as good as a standard blood count, and we say so.
Studies in which the signature beat chance. The signal is real, reproducible and explainable gene by gene.
PenetranSci is for research use only; it is not a diagnostic test. The full results, including what didn't work, are on the NeuroPenetrance evidence page.
Live app
See PenetranSci in action
A real, working application: patient intake with quality warnings, the six-gene panel, a report with a calibrated probability, a gene reference and batch analysis. Patient data is processed only in your session and never stored on the server.
What's next
Which patients carry the inflammatory form, and do they respond differently?
Across all eight studies, about one patient in five carries the same inflammation pattern. Next: checking it against measured CRP, working with biotech companies and research groups on their own samples and trial data, and preparing the first Engine 1 proof of concept.