Case Studies & Method Reports

Come see why we're excited about what we do and the future of AI-enabled drug development!


Finding the right patients, analogs, and indications for ONC201

ONC201 was a first-in-class anticancer compound in Phase II clinical trials, however its binding target and mechanism of action were unknown, which made future development and therapeutic positioning difficult. Our platform integrated data on ONC201's structure, efficacy, and genomic effects to identify and validate its mechanism of action. Based on the identified binding target, we ranked other imipridone analogs based on predicted selectivity and efficacy, and identified the optimal cancer subtypes for Oncoceutics to pursue clinically. These predictions were validated in vitro, in vivo, and through clinical trials.


Discovery of Novel Microtubule Inhibitors for Resistant Cancers

Microtubule inhibition therapy is one of the most successful forms of chemotherapy, however many patients become resistant to current therapies and thus there is a need to discover new compounds work on these resistant patients. Integrating data from over 7 different sources on over a million different compounds, we identified and validated a set of novel microtubule inhibiting (MT) compounds. We tested our top 4 predictions on cells derived from resistant patients and found that 3 new predicted compounds acted on resistant cancer cells. In fact, based on cytotoxicity assays, we found that our top prediction showed nearly 10x better efficacy in the resistant cell than the top performing approved treatment. 


Predicting Genotype Specific Synergistic Combinations

A major challenge in developing new combination therapies is the number of experiments that one has to do to identify context specific synergy. To address this we developed a new method to predict synergistic combinations for a given genotype or sample. Our platform integrates multiple features pertaining to a drug's structure, targets, effect as a single agent, and transcriptional response, and we observed how integrating these diverse data types led to a substantial increase in accuracy. We evaluated our method on a set of ~10,000 combinations tested on 60 different cancer cell lines and benchmarked an accuracy of over 90% at identifying cell line specific synergy. 


Repositioning Approved Drugs as Combinations for MT Resistance

In a clinical trial conducted at New York Presbyterian, we observed that approximately 50% of gastric cancer patients treated with docetaxol (taxol) presented with microtubule inhibitor drug resistance. Using genomic data from patients, our platform identified a unique signature that was correlated with resistance to taxol. We then mined through available approved drugs to pinpoint any that were predicted to reverse this signature (and thus bring resistant patients closer to a sensitive state). Based on this analysis we identified one candidate that was predicted to break resistance when combined with taxol. We observed a significant increase in overall efficacy in both the sensitive and resistance samples when they were treated with the newly identified combination, with overall cell survival dropping to ~30% in most samples.