AI in Drug Discovery and Development

Panel Discussion

Agenda

  • Panel Introductions
  • AI in Drug Discovery and Development
  • “Fireside Chat” (panelist Q&A)
  • Audience Q&A

Brian Repko (he/him)

  • 11 years at Novartis (now retired)
  • 8 years in oncology research
  • 5 years in an internal CRO (-omic biomarkers)
    • bulk RNA, DNA panel, scRNA, T/B cell, CITE-seq, spatial
    • including validation and qualification
  • Minneapolis-St Paul “chapter lead” for Bits-in-Bio
  • R/Pharma and {pharmaverse} contributor

Dr. Erika Kvikstad (she/her)

Geof Hannigan, PhD (he/him)

Director of Startup Engagement, Mayo Clinic

  • Passionate about making a positive impact on patient lives by blending science, technology, and entrepreneurship
  • Career spanning bench science, computational biology, business development, and startup engagement
  • Co-founder and former Executive Director of Merck Digital Sciences Studio startup accelerator program
  • Supporter of innovation ecosystems, including founding the Minnesota biotech networking group BiotechMN
  • Blogging at Living Bio (https://livingbio.substack.com)
  • Email: hannigan.geof@gmail.com

What is Drug Discovery and Development?

AI and DD&D - Brian (1/2)

  • AI has always been a part of DD&D - comp.bio and comp.chem as fields
    • New focus on GenAI, agentic workflows or skills, and improved tools
  • Novartis efforts - data42, scientific vs operational AI, tooling
    • Huge efforts to collect and connect preclinical, clinical trial, and real-world data
    • Investments in predictive safety/tox, therapeutic design, and lab automation
  • Data fit, quality, and integration are key to model success
    • Unresolvable gaps in data - missing / mistaken / methods
    • Existing knowledge is sometimes not reproducible or not accurate
  • Hype is a bit out of control (Anthropic S-1) and DD&D is a miracle
    • We are still learning basic biology and developing new methods (shRNA/sgRNA)
    • Genuine effort on where to set the bar here

AI and DD&D - Brian (2/2)

  • Drug Development / GXP
    • CDISC efforts - 360i, USDM / TMF, SDTM-ADAM and data engineering coding
    • Only non-critical tasks and then with auditing and approvals
    • Validation of agents similar to that of R and R packages (with non-determinism)
  • Regulatory Issues
    • FDA already using AI for review of protocols and trial data (ELSA)
    • Sponsors learning how to do the same on their side prior to submission
  • What is successful use of AI in DD&D? What is ROI on the tools?
    • Somewhat timelines (and perhaps cost) - biological latency still exists
      • If we lower cost, does that change what we go after?
    • Success as improving p(success) - better models, biomarkers, and trial designs
      • If more drug candidates, do we create a bottleneck with patients? sites?

AI and DD&D - Erika (1/2)

We can build a better drug discovery process

AI and DD&D - Erika (2/2)

Two questions on the path to better medicines

AI and DD&D - Geof (1/2)

Leveraging AI Across the Pipeline

AI and DD&D - Geof (2/2)

AI Trends in Drug Discovery & Development

Fireside Chat

Audience Q&A

Thank you!