Insights

2023 was a defining chapter in pharmaceutical innovation, with artificial intelligence emerging as the driving force in drug discovery and development. The convergence of cutting-edge technologies and collaborative endeavors is reshaping the future of medicine, resulting in breakthroughs and insights that hold real promise for the year ahead. As the year draws to a close, now is a good time to reflect on the key milestones that have marked the intersection of AI and pharma in 2023.

Regulatory Acceptance as FDA Embraces AI

In a pivotal move, 2023 saw the FDA accept the role of AI and machine learning in potentially transforming drug research and development, biological products and medical devices, issuing a discussion paperto engage stakeholders and explore AI/ML use in these areas. With a surge of submissions incorporating AI and ML components, the FDA’s acknowledgement of the need for an adaptive regulatory ecosystem marks a crucial step toward integrating these new technologies into the fabric of pharmaceutical development.

NVIDIA Launches Cloud Offering for Precision Drug Development

In March, AI trailblazer NVIDIA launched its BioNeMo™ Cloud service offering, providing domain-specific generative AI models for drug development that enable researchers to fine-tune and customize their models for optimal results. BioNeMo has already seen significant adoption, with pharma giant AstraZeneca and biotech startup Evozyne using the new cloud service to develop “supernatural” proteins that are potentially more effective in accelerating drug development than naturally occurring proteins.

MIT’s DiffDock Represents a Game-Changer in Molecular Docking

Also in March, researchers at MIT introduced DiffDock, a molecular docking model for faster drug discovery. Based on diffusion generative models, DiffDock represents a gamechanger in molecular docking prediction. By mapping the degrees of freedom associated with docking, DiffDock boasts a success rate of 38%, surpassing traditional methods by a significant margin.

Phase II Success for Insilico’s Generative AI

Insilico Medicine’s generative AI tool InClinico identified INS018_055, the first AI-discovered drug entering Phase II trials for idiopathic pulmonary fibrosis. Trained on a vast dataset of 55,600 trials, InClinico has achieved an unprecedented 79% accuracy in predicting outcomes of real-world clinical trials. Demonstrating substantial cost and time savings (at a tenth of the cost and a third of the time compared to traditional methods), this has highlighted the potential of AI in expediting drug development timelines.

DeepMind’s AlphaMissense Predicts Genetic Mutations

Unveiled in September, DeepMind’s AlphaMissense AI toolanalyzes genetic mutations to predict whether genetic mutations are potentially pathogenic. Achieving a 90% accuracy in predicting genetic disease, it has the potential to accelerate research into rare diseases.

Absci Achieves Breakthrough in De Novo Drug Creation

Biotech research company Absci announced a breakthrough in generative AI drug creation, becoming the first to design and validate de novo therapeutic antibodies with zero-shot generative AI using no training data. With a hit rate up to five to 30 times greater than biological baselines, this could potentially accelerate drug candidate timelines and improve success rates clinical trials.

A Year of Collaborative Ventures

2023 proved to be a year of collaborative endeavors, with key players joining forces to accelerate drug discovery and development:

• Ginkgo and Google Cloud joined forcesto deploy large language models via Vertex AI for genomics, protein function, and synthetic biology, with the aim of expediting drug discovery, agriculture, and biosecurity advancements.
• Fujitsu partnered with RIKENto develop an AI drug discovery technology using generative AI to predict protein structural changes from electron microscope images, reducing the development time and cost of drug discovery.
• Aiming to transform patients’ lives through the power of AI, Insmed and Google Cloud collaborated on a generative AI modelto accelerate drug discovery, reduce timelines and drive the development medicines for rare diseases.
• AstraZeneca partnered with Verge Genomics, utilizing the latter’s CONVERGE™ machine learning platform for novel drug target discovery in rare neurodegenerative and neuromuscular diseases.
• Merck KGaA extended its AI capabilities through strategic collaborations with BenevolentAI and Exscientia, focused on enhancing drug discovery in oncology, neurology, and immunology.
• NVIDIA invested $50 million in Recursion Pharmaceuticalsto support the training of the company’s AI models for drug discovery, with plans to license the models to biotech firms via BioNeMo.

AION Labs: Exploring the Future of Drug Discovery

As we reflect on a transformative year for AI-powered pharmaceutical innovation, AION Labs invites you to become part of our vibrant community that is reshaping the future of AI in drug discovery and development. If you’re interested in learning more, or just spotted something I may have overlooked, drop a comment below and don’t hesitate to reach out and connect.

By Arnon Fluksman, PhD, Scientific Analyst, AION Labs