Biologics are complex and varied, making it hard to predict their developability and immunogenicity. Simply put, we need more extensive and diverse data to train better predictive models. However, the high costs of generating data for certain properties, like high-concentration viscosity measurements, mean that no single company has all the necessary data to produce highly accurate models. Federated learning (FL) offers a solution by enabling collaborative model training across multiple organizations without the need to share sensitive data. Active Learning, in turn, enables cost sharing on the acquisition of new data.

The FAITE consortium, will leverage federated and active learning to train models for predicting properties of biologics. It addresses data privacy challenges and enhances model performance, thereby reducing discovery costs and minimizing development risks by avoiding designs prone to late-stage failures. The consortium’s collaborative nature fosters innovation and optimizes resources, leading to the development of safer, more effective biologic therapies. This consortium positions the participating companies at the forefront of AI-driven drug discovery, significantly benefiting patient care and public health.