Schedule

Drug discovery efficiency is key to deliver life-saving medications to patients in a cost-effective manner. High-throughput chemistry platforms are increasingly becoming a part of modern Medicinal Chemistry departments due to the strategic advantages they offer in improving drug discovery efficiencies. Herein, we will introduce the audience to the recent evolution of new design principles in medicinal chemistry and culminate with our recent work on merging synthetic chemistry, high-density experimentation and biological assays – a concept that we term direct-to-biology (D2B).

Highlighting how the development of sustainable enzyme-based catalysis for large-scale manufacture of APIs has evolved at Pfizer. Emphasis on strategies to overcoming bottlenecks towards achieving critical process targets, and the development of technologies enabling faster enzyme discovery, engineering, and process optimization. Also discussed, prominent examples for the commercial synthesis for APIs, illustrating implementation to >1 kg and the power of collaborative research in achieving the goal of large-scale manufacture.

The integration of novel reaction methodologies from academic research into drug discovery processes is frequently impeded by inadequate characterization of substrate scope, particularly when dealing with complex bioactive precursors. This presentation will elucidate how our High-Throughput Experimentation (HTE) platform has been effectively deployed to incorporate contemporary transformations for the synthesis of novel analogs, thereby advancing medicinal chemistry research. Additionally, we will illustrate how HTE serves as a pivotal bridge between academia and industry, fostering the establishment of modern reaction development protocols and significantly catalyzing innovation.

Recent advances in machine learning (ML), notably large language models, rely heavily on extremely large datasets. Applying ML to chemistry and chemical synthesis has significant potential for improving chemical design and reaction planning. For instance, accurate yield prediction could optimize reaction conditions and reagent selection a priori. However, datasets in this area are typically small, limited by low-throughput liquid chromatography–mass spectrometry workflows, specifically due to the separation step. Mass spectrometry itself offers high sensitivity and throughput, presenting an opportunity for substantial improvement. Here, we introduce a massive chemical reaction dataset generated using flow injection mass spectrometry, aimed at enabling the training of next-generation foundation models for chemistry. We collected extensive data on imine synthesis reactions, including detailed sample tracking, kinetic analysis capability through multiple data points, and dilution studies to examine mass spectrometry signal intensity relative to concentration. The dataset comprises information on over one million sample-compounds, 794,928 of which possess unique mass.

Automation and high-throughput (including parallel) experimentation offers an attractive way to
generate large amounts of data. This data can be useful on its own in a screening context or useful for training predictive models able to extract useful insights. I will talk about our own lab’s path toward an automated laboratory that complements our primary work in AI for chemistry. I will discuss two key vignettes: one uses HTE to investigate polymer blends to achieve emergent function not achieved by any constituent polymer as an example of formulation optimization; one uses HTE to screen many reaction conditions for the purposes of identifying novel combinations of known mechanistic pathways for multicomponent reaction discovery. In both cases, I will discuss how the scientific goals must be mapped onto practical workflows accessible given the hardware and software constraints of our academic laboratory.

The use of parallel microscale screening to identify optimal reaction conditions is well-appreciated. However, numerous technological and reproducibility hurdles limit the types of data sets that can be generated and their fidelity. Approaches to overcome these limitations will be discussed.

High Throughput Experimentation (HTE) is a versatile and powerful technique for the development
of new metal-catalyzed reactions, particularly asymmetric organometallic reactions. In such cases, extensive screening of ligand, metal, substrate, and solvent may be required to develop an efficient, selective, and general reaction. Such campaigns are slowed and complicated by the complexity of analysis and sourcing the necessary ligand libraries. This talk will detail efforts to optimize HTE workflows for the development of new asymmetric organometallic reactions through the development of ligand library infrastructure and analytical workflows and their application in several recent reaction development campaigns.

Large arrays of rationally designed experiments are a powerful tool for solving complex and challenging chemical problems. Our chemical high-throughput experimentation journey will be presented in several dimensions, from bespoke prototypes to commercially available tools, from simple manual methods to complex automated experiments, and from microscale to nanoscale experimentation. Conceptual and practical aspects of high-throughput experimentation will highlight the broad applicability of these tools in chemical research.