Client success story
ML Driven Protein Complex Predictive Modeling
Business Value: Candidate Molecules Identified Quickly
Summary: AI-Driven Improvements To Molecule Identification
A biotechnology company was developing small-molecule therapeutics that target protein complexes, but the process to identify and model higher-order complexes was time-consuming, highly dependent on both wet lab resources and human intuition, and as a result, limited the client’s ability to develop new therapeutics efficiently.
Challenge: Modeling Tools Required Customization And Validation
Prediction of small molecule and protein complexes has been significantly accelerated by Generative AI, but the extant open source and commercial packages were not suitable for the client’s purpose, which involved modeling complex physical phenomena and integrating datasets with radically different qualities and sample counts. The client needed a simulation and small-molecule design capability that could leverage public and private data resources to accelerate drug development.
Solution: Development Of Protein Complex And Binding Prediction Tools
The bPrescient team worked with the client to develop models for complex prediction using generative AI and input data from several heterogeneous assays measuring distinct properties of each small molecule related to predicted efficacy in subsequent phases of testing. bPrescient implemented a pipeline composed of open-source and proprietary AI and physics-based models, which collectively improved the client’s ability to rapidly design new lead molecules and understand those design principles and how they relate to protein structure and complex formation.
Outcome: New Candidate Molecules Identified
The client was able to apply the methods developed to predict complex formation and therapeutic efficacy, leading to new therapeutic molecules to test.
Let’s talk
Tell us about your toughest R&D challenge and let’s discuss how bPrescient can solve it.
