During the drug discovery process, it is standard practice to test a large library of compounds for their biological activity towards both the intended target molecule as well as peripheral off-target molecules. This process can be very time- and money-consuming. Pharmaceutical organizations are now accelerating this process by employing predictive modeling approaches that are used to generate quantitative predictions of molecular activities in a fraction of the time. In this talk, we will discuss Vertex's approach to building such a predictive modeling infrastructure using modern machine learning operations and management techniques. We have built a modern MLOps platform using AWS-native services such as Amazon SageMaker, AWS Batch, AWS Lambda, API Gateway, S3, and Fargate which are designed to deliver low latency predictions, separate the model management from model training concerns, and provide well-defined interfaces for various applications and users. This platform enabled us to train, deploy, and manage tens of thousands of activity prediction models which can be retrained daily or on-demand. We also provide a scalable, flexible model management experimentation infrastructure that provides a bedrock for collaboration and ultimately delivers optimized processes for testing small molecules.

In this presentation, we will be exploring the use of generative AI in research based on a use case for the Dutch Healthcare Institute. Everyone recognizes the capabilities of generative AI these days, but how can you use this in a reliable setting? A setting where the outcome of generative AI is part of the policy-making process or research.

Traditional drug discovery is known for its high costs, lengthy timelines, and frequent lack of success. Machine learning has the potential to positively impact any or all of these aspects, namely cost, speed, and success. I have the privilege of leading a global team at AbbVie that specializes in the application of machine learning to drug discovery. During this talk, I will delve into how AbbVie is addressing the challenges of drug discovery through the utilization of machine learning. I will share various use cases, both successful and unsuccessful, shedding light on what the future holds for the pharmaceutical industry. Listeners will gain insights into how to effectively scale drug discovery using machine learning and will discover actionable opportunities that can benefit us all. The presentation will provide a glimpse into the strategies and algorithms of machine learning currently in use in drug discovery, offering a forward-looking perspective on the direction in which our collective efforts are headed.

Foundation models (FMs) have catalyzed incredible AI breakthroughs in natural language. How far are we from achieving something similar for the biological language that connects DNA, proteins, cells, tissues, and organisms? In this talk, we unveil Bioptimus’ plan to bring together biological data to build a universal FM for biology—unleashing the full power of machine learning to transform our understanding of the living world, disease, and medicine.

Generative neural networks have begun to receive immense attention in the field of pharmaceutical development due to their ability to interpret human-readable representations of data, such as plain-text experimental methods and SMILES strings. This makes them suitable for answering a wide variety of questions across many domains. In this talk, we discuss how such a model has been used to address a topical question in a novel way. A specific example will be shared of a transformer neural network being used to create a model with potential to accelerate the development and de-risking of pharmaceutical solid forms.

Some mammals have naturally evolved strategies to prevent or reverse disease pathologies common in humans, including the ability to repair fibrotic scar tissue, drastically alter metabolic rate, and regrow neurons. At Fauna Bio, we use machine learning approaches, our unique dataset, and our custom Graph Neural Network (GNN) trained on our proprietary biomedical knowledge graph to design therapies that allow humans to tap into the protective and regenerative strategies from 100 million years of mammal evolution. Learn about nonconventional approaches to drug discovery and development, utilizing ML-powered strategies.