Convening innovators from industry, academia, and government

By Maeve Smillie

Despite a 44% increase in drug development investment over the past decade, the pharmaceutical pipeline has slowed, yielding fewer new drugs at greater cost, with only limited efficacy. As the industry grapples with this paradox, the Rafik B. Hariri Institute for Computing’s hosted the AI for Drug Discovery Open Innovation Forum  on Wednesday, October 30, 2024, bringing together more than 160 innovators from industry, academia, and government to confront these challenges head-on. New developments in AI hold the promise to address the challenges which will be made possible through collaboration.

The day-long event consisted of keynotes and panels which highlighted state-of-the-art AI technologies and allowed participants to work through pressing challenges in drug discovery. Attendees split up into breakout sessions to pin-point and prioritize key AI technologies, pressing challenges, and activities that should be developed collaboratively and in open-source environments.  The event also featured a hands-on session with opportunities to work with open-source models such as Biomedical Foundation Models, which are fundamental for drug discovery. The forum served as the kickoff event for a new AI for Drug Discovery working group with the AI Alliance. 

“The forum was the outcome of a unique partnership between the Hariri Institute at BU, IBM Research, the Cleveland Clinic, and the Hariri-based Mass Open Cloud Alliance,” Yannis Paschalidis, Director of the Hariri Institute for Computing noted after the event. He added that the event “illustrated the immense interest for AI and its expected large impact and footprint in the biotech industry. The use of AI for research and development in biotech is quite advanced and beyond experimental. It was refreshing to see the interest in open-source models and the value participants see in such an approach (in terms of model quality, diversity of data sets used, and impact).” 

Welcoming Remarks

Gloria Waters, PhD, Provost and Chief Academic Officer at Boston University, kicked off the forum by welcoming an esteemed audience of scholars and industry leaders. She then introduced officials from Massachusetts state government to share their opening remarks. Yvonne Hao, Secretary of the Executive Office of Economic Development, and Kirk Taylor, President and CEO of the Massachusetts Life Sciences Center, highlighted the state’s commitment to innovation, including a 10-year reinvestment in the state’s life sciences sector. Hao said “We looked at each different area where we have big problems we want to solve and the ways that AI can accelerate that. We want to make Massachusetts a leading hub for AI.” Later she explained that more than $1 billion of the Mass Leads Act –  a $4 billion economic development bill signed into law by Governor Healey on November 20 – will go toward life sciences and AI initiatives that will continue to grow Massachusetts’ lead in the area of AI for drug discovery.

Following their remarks, Yannis Paschalidis, Hariri Director, and Jianying Hu, IBM Fellow and Global Science Leader in AI for Health, underscored the importance of collaborative efforts in addressing challenges in drug development.

The opening keynote presenter, Joseph Loscalzo, delved into the high failure rates of traditional empirical drug discovery methods and explored emerging approaches aimed at improving success rates. He emphasized the potential of network-based, AI-driven, and quantum-inspired strategies to revolutionize how diseases and drug targets are identified. The forum’s central mission was to spotlight the critical challenges in drug discovery and foster interdisciplinary collaboration to develop innovative solutions. Additional keynote speakers included Jeremy Jenkins, US Head Discovery Sciences, Novartis BioMedical Research who gave a talk entitled ““Generative AI Applications in Early Drug Discovery”;  Sean Mooney, Director, Center for Information Technology, NIH, whose talk focused on “AI@NIH: Drug Discovery, Clinical Healthcare, and Multimodal AI”; as well as Princeton University Computer Science Professor Mona Singh.

Panels 

There were two panel discussions during the forum, each tackling critical aspects of AI’s role in drug discovery. The first panel, moderated by Diane Joseph-McCarthy, focused on Drug Discovery Bottlenecks and How AI Can Help. This discussion brought together speakers from medicine and data science, and the tech industry, who highlighted key challenges in the drug discovery pipeline, such as lengthy development timelines, high costs, and the complexity of analyzing vast datasets. The panel explored how AI tools could streamline these processes, helping researchers identify viable drug candidates faster and with greater precision. The speakers also emphasized the need for integrating domain-specific expertise with AI technologies to address these bottlenecks effectively. The second panel, New Generative AI Approaches for Drug Discovery and the Need for Open Innovation, was led by Michal Rozen-Zvi and featured panelists from journalism, medicine, and technology.

This discussion highlighted how generative AI models, such as those capable of designing novel molecules or predicting protein structures, are transforming drug discovery. The panel emphasized the importance of open innovation, encouraging collaboration across disciplines and the sharing of models and data to accelerate progress. They also discussed ensuring transparency, reproducibility, and accessibility in generative AI research.

Breakout sessions

After the panels, participants divided into breakout sessions, each focusing on key topics such as Benchmarks, Multi-Modal Fusion, Model Architecture, and Challenges and Competitions. These sessions were designed to encourage open dialogue and collaboration across disciplines, bringing together experts from academia, industry, and research institutions.

By fostering interdisciplinary conversations, the breakout groups created a space where attendees could share insights, identify common challenges, and explore innovative approaches to advancing drug discovery. The hour-long discussions allowed participants to focus on their selected topics, blending diverse perspectives and expertise to tackle complex problems. At the conclusion of each session, group leaders presented their findings, synthesizing ideas and proposals that could guide future work in AI-driven drug discovery. These breakout groups played a critical role in the forum’s success by emphasizing the importance of collaborative innovation. The diversity of expertise within each group  ensured that solutions were not siloed within a single field but instead reflected a comprehensive understanding of the challenges and opportunities in the development of AI models for drug discovery. 

Purpose of the Event

The forum also featured a hands-on session where participants worked with pre-deployed AI models by IBM, using Red Hat OpenShift on the Mass Open Cloud (MOC). Participants tested models in an efficient, interactive setup, showcasing how OpenShift AI streamlines building, testing, and scaling models for scientific research, specific to drug discovery. Participants explored protein combinations, learned about training open-source models, and utilized MOC’s open-access cloud platform, which further fosters collaboration between academia and industry. With MOC accounts, users can join the Drug Discovery project, accessing the same tools to advance AI-driven drug discovery.

The purpose of both the breakout and hands-on sessions was to bring together participants with diverse skill sets and knowledge bases to foster connections and build a foundation for future collaboration in AI-driven drug discovery. By combining expertise from fields like biology, chemistry, data science, and computer engineering, and politics, the sessions encouraged interdisciplinary problem-solving and innovation. These interactions aimed to break down silos, ensuring that the development of AI models for drug discovery benefits from a wide range of perspectives, ultimately driving more effective and ethically sound solutions.

Co-sponsors of the event included: Boston University’s Hariri Institute, BU Bioengineering Technology & Entrepreneurship Center, MOC Alliance, Red Hat, IBM, the Cleveland Clinic and the AI Alliance.

To join the AI Alliance Mailing List, click here!