Core Insights - Predictive Oncology Inc. has developed predictive models from 21 unique compounds sourced from the Natural Products Discovery Core at the University of Michigan, marking a significant advancement in AI-driven drug discovery [1][2] Group 1: Drug Discovery and Development - The predictive models were created using an active machine learning platform that aims to reduce the time required for drug candidate selection while enhancing the likelihood of technical success [2] - The Natural Products Discovery Core houses one of the largest pharmaceutically viable natural products libraries in the U.S., with specimens collected from diverse global regions [3] - Natural products have been crucial in drug discovery, with at least 50% of small-molecule drugs approved in the last three decades derived from these compounds [4] Group 2: Testing and Results - Three compounds showed strong tumor drug responses across all tested tumor types, outperforming Doxorubicin, a benchmark anti-cancer drug [5] - The predictive machine learning model was able to make confident predictions covering 73% of all experiments after only 7% of possible wet lab experiments were conducted, potentially saving up to two years of laboratory testing [6] - Initial results were achieved using only about 1% of the available NPDC library, indicating significant potential for further exploration of additional compounds [7] Group 3: Company Overview - Predictive Oncology utilizes a scientifically validated AI platform, PEDAL, which predicts with 92% accuracy whether a tumor sample will respond to a specific drug compound [8] - The company maintains a biobank of over 150,000 assay-capable heterogeneous human tumor samples, providing extensive resources for drug discovery [8]
Predictive Oncology Successfully Develops Predictive Models Derived from Never-Before-Seen Compounds for Prevalent Cancer Indications Including Breast, Colon and Ovary