OneThree's Workflow for Predicting Drug Toxicity

Background

OneThree Biotech was born out of the largest precision medicine institute in America with the purpose of decreasing the failure rate of bringing therapeutics to market using biology driven artificial intelligence. Our technology was builtin a lab setting; allowing us to engineer our methodologies around real-world results. While working with major biotech and pharma partners; we’ve validated our predictions through work in target discovery, toxicity prediction, biomarker selection, and much more. In this white paper, we’ll discuss the OneThree Adverse Event module and how it was used to identify safe version of toxic drugs and pinpoint the MoA driving toxicity for a clinical stage compound.

Problem

Adverse events account for a large portion of clinical trials failure, which end up being costly and dangerous. 17% of Phase 3 clinical trial failures are due to adverse events, even though safety is specifically evaluated both pre-clinically and in earlier trials. Even worse, numerous drugs make it to market and are then found to cause severe adverse reactions leading to costly withdrawals. Identifying a drug’s toxicity pre-clinically would enable drug developers to avoid clinical trial failures and drug withdrawals.

Adverse Event Prediction

Ranging from cardio to neurotoxicity, OTB’s Adverse Event module predicts tissue specific adverse events for a given small molecule. This module leverages commonly available pre-clinical data (cell viability screens, compound structure, etc.) and has benchmarked significant predictive performance across seven common adverse events (AUC = >0.8). When tested, this module accurately identified 80% of drugs previously withdrawn for adverse events, using only data available pre-clinically and was able to accurately separate approve drugs from those that failed earlier due to toxicity reasons. The interpretability of this module allows for researchers to identify if toxicity is due to chemical structure of target selectivity, which can be used to inform future drug development. 

Analog Selection

OTB’s Adverse Event module enables rapid assessment of entire compound libraries to determine the safest compounds to move forward with for development. This module easily fits into early R&D pipelines and can be used to refine analogs for further development and filter out potentially high risk compounds. For example, Actos, a type 2 diabetes medication, was developed after its analog, Avandia, had to be recalled within the US for cardiotoxicity. The OTB Adverse Event Prediction module predicted significantly higher cardiotoxicity for Avandia versus Actos and if it were implemented pre-clinically, could have helped avoid the costly development of Avandia. 

Predicting Tissue Specific Effects of Mivebresib

The OTB Adverse Event module can identify the source of the predicted toxicity, allowing for more informed drug development decisions. To evaluate this, we applied to, Mivebresib, a pan-BET inhibitor in clinical development for various cancer indications. We predicted a high likelihood of Mivebresib resulting in blood based adverse events, which lined up with the report that 56.9% of patients within Phase 1 clinical trials reported suffering from thrombocytopenia. When interrogating the model results, it was shown that the high toxicity prediction was driven by Mivebresib pan-BET inhibition, and the likelihood of blood toxicity significantly decreased if this compound were reformulated to be a selective BET inhibitor. This finding was then validated through many independent reports which proposed that pan-BET inhibition was a driver of adverse events. Overall these findings showed how the OTB Adverse Event module could be used to predict adverse events and identify the underlying mechanism of any predicted toxicity.

Relevant Publications:

  1. A Data-Driven Approach to Predicting Successes and Failures of Clinical Trials
  2. A Machine Learning Approach Predicts Tissue-Specific Drug Adverse Events