(699b) Accelerated Development of Active Pharmaceutical Ingredients with Machine Learning Driven Automated Platforms

The lab-scale development of API or intermediate synthesis involves optimization of reaction conditions such as reaction temperature, residence time, and stoichiometry, which are traditionally achieved with multiple experiments with one factor at a time (OFAT) approach. We present multiple case studies at Pfizer Chemical Research and Development (CRD), Groton CT site employing this autonomous self-optimization platforms to enable the identification of optimal conditions for synthesis of APIs, while reducing the amounts of raw materials consumed, compared to OFAT approach. This automated platform requires minimal human intervention, relieving expert scientists of manual tasks so that they may focus on new ideas. We present a development of PFR platforms with an in-house developed automation software and a scheduling algorithm for orchestrating all hardware operations. We integrate automation software with equipment, in-line process analytical technology (PAT), and iterative experimental design (DoE) based on artificial intelligence (AI) and active machine learning (ML) optimization algorithms. We present the comparative performances of Bayesian Optimization models for data-rich screening including Gaussian Process, Deep Ensemble, Gradient Boosted Discission Trees, and Dynamic for reaction optimization campaign over a wide range of chemical domains.