Workshop Recap: AI-Powered Advances in Small-Molecule Drug Discovery
Date: November 8, 2025 @ 12:00 am –
Location: 1633 Old Bayshore Hwy #280, Burlingame, CA 94010
Organizer: CABS STC | Sponsor: HitChem
Overview
On October 25, 2025, the Science & Technology Committee (STC) of Chinese American Biopharmaceutical Society (CABS) hosted a half-day workshop titled “Smarter Molecules, Faster Cures: AI-Powered Advances in Small-Molecule Drug Discovery” in Burlingame, California. The event attracted more than 90 scientists, entrepreneurs, and AI innovators from academia, biotechnology, and pharmaceutical industries. Attendees explored how artificial intelligence is transforming every step of small-molecule discovery—from generative chemistry to toxicity prediction and high-throughput screening.
The workshop was co-chaired by Dr. Liping Meng (Gilead, CABS STC Co-Chair) and Dr. Alex Yang (Vir Biotechnology, CABS STC Co-Chair), who opened the session by emphasizing AI’s potential to reshape the traditional paradigm of drug discovery.
Scientific Presentations
● Dr. Wenhao Gao (Stanford University) – Navigating Synthesizable Chemical Space with Generative Modeling
Dr. Gao presented state-of-the-art generative AI frameworks for designing synthetically accessible molecules. He demonstrated how incorporating reaction knowledge and synthesis constraints enables AI to propose compounds that are both novel and experimentally feasible.
● Dr. Jin Wang (Baylor College of Medicine) – Powering Next-Gen Therapeutics with High-Throughput Proteomics
Dr. Wang showcased next-generation proteomics platforms that accelerate the discovery of molecular glues and covalent drugs. His integrated workflow—combining high-throughput docking, biochemical assays, and animal models—illustrated how multidisciplinary approaches can streamline the path from hits to therapeutic leads.
● Dr. Yuxing Peng (NVIDIA) – Bridging Physics and AI: Hybrid Approaches for Accelerated Drug Discovery
Dr. Peng introduced NVIDIA’s hybrid AI-physics computational pipeline, integrating molecular-dynamics simulations with machine-learning models to enhance docking accuracy and free-energy prediction. He emphasized the importance of generating high-quality experimental and simulated data, noting that AI does not replace physics—rather, it helps us understand physics better and leverage human expertise in the process.
- Dr. Zhe Wu (Exelixis) – The Era of Machine-Integrated Drug Discovery
After the coffee break, Dr. Zhe Wu, Director of Computer-Aided Drug Discovery at Exelixis, gave an insightful presentation titled 'The Era of Machine-Integrated Drug Discovery'. He described how the drug discovery process is shifting towards a machine-integrated approach. Rather than following the traditional "Design–Make–Test" loop, teams now incorporate a "Learn" step that employs physics and machine learning to improve design and synthesis decisions. Dr. Wu also presented Boltz-2, a recent advance in accurate and efficient binding-affinity prediction which improves hit identification by rescoring docking results and guiding downstream free energy perturbation (FEP). Together, AI/ML models and physics-based methods accelerate discovery and help scientists prioritise what to make and test. The key takeaways are: adopt a predict-first culture to raise hit rates, reduce wasted cycles and make data-driven decisions.
● Dr. Alejandra Trejo-Martin (Gilead Sciences) – In Silico Screening and Toxicology Assessment of Impurities
Instead of focusing on AI, Dr. Alejandra Trejo-Martin, Senior Research Associate at Gilead Sciences, presented “Toxicological Assessment of Impurities.” She defined impurities as any component of a product that is not the intended drug substance (API) or an excipient, and reviewed common sources arising from manufacturing and storage. She outlined the major impurity types and the corresponding regulatory guidelines, and walked through impurity-qualification strategies and study designs in detail. Dr. Trejo-Martin stressed the importance of assessing and controlling impurities early in development. While this workflow is not yet powered by AI/ML, in-silico screening and risk assessment could be strengthened by a machine-integrated approach in the future.
● Dr. Hongbo Zhang (HitChem) – Leveraging Generative AI to Develop CRBN Molecular-Glue Libraries
In the final talk, Dr. Hongbo Zhang, VP, head of drug discovery service at Hit Chem, who is also the major sponsor of the workshop, presented "Leveraging Generative AI to Develop CRBN MG Library for HTS and Rational Design". He started with a breif overview of HitChem's chemisrty service and recent publications. Then walked through the workflow in detail, including aligning fragment-aware molecular representations with LLMs, addressing model bias from training data through fine-tuning and chemist QC, and deploying the system as a practical assistant for medicinal chemists. Case studies showed that the HitChem AI model can generate diverse, novel scaffolds with drug-like properties that are synthesizable in few steps, and that the focused CRBN library improves hit rates in screening compared with off-the-shelf collections.
Panel Discussion and Networking
The closing panel and Q&A, moderated by Dr. Liping Meng Principle Scientist II at Giliead and Co-Chair of the CABS Science & Technology Committee, highlighted the key challenges facing AI-driven drug discovery — including data quality, the high cost of labeling, and the urgent need for a pre-competitive data-sharing ecosystem. Panelists also stressed the importance of rigorous wet-lab validation for both datasets and models, noting that legal and regulatory teams remain cautious about AI outputs without human oversight. The discussion compared physics-based and data-driven methods and agreed that hybrid workflows often yield the best results: AI can automate tedious, repetitive tasks, while physics keeps designs mechanistically grounded. Participants emphasized that dataset harmonization remains a major hurdle and that choosing the right biological target often matters more than designing the cleverest molecule. Bridging the gap between AI algorithms and experimental validation—and fostering academic–industry collaboration—will be critical for future progress. The battle is not between technologies or between tools and humans, but about how we expand our knowledge into the unknown.
Closing Remarks
The event concluded with warm applause and enthusiastic feedback. During the networking lunch, participants exchanged ideas and forged new collaborations in a collegial atmosphere. The 2025 CABS AI Workshop successfully delivered what the community cares about—connecting AI innovation with translational science—and left attendees inspired to tackle the scientific and ethical challenges ahead.
Dr. Liping Meng closed the session by thanking all speakers, sponsors, and volunteers, reaffirming CABS’s mission to foster collaboration and scientific innovation across the global biopharmaceutical community. The workshop underscored how AI-driven chemistry is transforming drug discovery—turning data into design, and design into cures. “From data to design, AI is helping us build smarter molecules—and faster cures.”
This event was proudly sponsored by HitChem, whose generous support helped make this workshop possible.
HitChem has been focusing on identifying hit compounds for novel targets since 2018. The team has successfully helped clients discover many early-stage candidates, utilizing high-quality libraries for HTS, unique HTVS workflows, and 2D molecule generation model.
HitChem provides comprehensive chemistry services and custom compound libraries, including molecular glue, highly diverse libraries, CNS, cyclic peptide and covalent libraries. In the dry lab, HitChem offers CADD, MD simulations, and molecular generative models.
.