Evidence
Predicting how real biology responds to drugs, without training data.
Whole-cell AI clones reproduced wet-lab outcomes across cancer and autoimmune diseases. Because the platform learns nothing from data, every case below was new to the system.
“The infinitusbio's Human AI Cell Clone Foundry has enabled us to validate our wet-lab and patient findings across two cancer types while uncovering novel combination opportunities we hadn't previously seen. In our experience, unlike other AI approaches, their native multi-agent system needs no training data. This eliminates historical data bias and enables the discovery of new cell–drug interaction mechanisms where data is absent.”
“We're delighted to collaborate and advance our next generation immunotherapy programs with the support of infinitusbio's Digital Cell Clone platform. The partnership follows infinitusbio's successful Digital Cell Clone simulations using Oncovita's vaccine candidates, which confirmed the wet-lab results generated by Oncovita.”
“This will revolutionize genomics because we can now model the genome and obtain answers in a way that is otherwise impractical today, with current technologies.”
Cancer Cell-line Ranking
Ranking 17 lung cancer cell lines against the wet lab, with no training.
The question. Which of 17 small-cell lung cancer (SCLC) cell lines respond best to this drug?
What we used. mRNA-seq of each cell line and the drug's on- and off-targets. No prior response data.
The result. Our ranking correlated with the wet-lab GI50 ranking at Spearman ρ = 0.82, p = 0.00005.
Spearman correlation with the wet-lab ranking
ρ 0.82
No machine-learning benchmark is directly comparable. Those methods must first be trained on historical response data, and that training carries the biases of the data into every prediction. infinitusbio is not trained, so the result rests on the mechanism alone.
The same approach held across seven validations in cancer and autoimmunity.
Two unrelated domains: cancer drug response, and inflammatory (IBD) macrophage signalling. 88–100% agreement with wet-lab outcomes on every classification test.
Agreement between whole-cell AI clones and wet-lab outcomes
Partners validated our results in wet-lab and patient studies.
Found a novel target
Identified a novel IBD target for patients resistant to anti-TNF (Humira). Client-validated in months.
Predicted the failure
Predicted a major oncology drug would fail. The sponsor later terminated the trial after Phase II.
Found a new indication
Confirmed the primary indication, correctly grouped by cancer type, and validated a new treatment plan in ovarian cancer.
Novel combinations
Confirmed the PKMYT1i combination as the best among small molecules, and predicted the best combination overall with a PD-L1 inhibitor.
More results are available under NDA.
Including non-public combination data we can demonstrate on your drug within weeks.