AI for Science in San Diego
We're gathering a group of researchers, engineers, and scientists in La Jolla to explore how AI is accelerating discovery across fields like biology and chemistry.
Expect talks on cutting edge applications of ai models, and candid conversations about where AI-driven science is headed next. Food, drinks, and venue provided!
Agenda
6:00 - Doors open. Drinks + mingling
6:15 - Talks, session one (three talks)
7:15 - Brief intermission
7:30 - Talks, session two (three talks)
8:15 - Open discussion + mingling
9:30 - Wind down
Speakers
Rose Yu | UC San Diego
Guiding generative AI with molecular dynamics to design molecules with real activity
Sural Ranamukhaarachchi | Serafin Labs
AI on "metastasis-in-a-dish" organoid screens to find new cancer therapies
Manoj Hariharan | Salk Institute
Reading DNA methylation across human tissues
Emily Manoogian | Salk Institute
Your Body Is a Time Series: AI, Circadian Rhythms, and Precision Health
Charlie Bowman | Ward Lab, Scripps Research
Wiring AI into a Structural Biology Lab: Research, Operations, and Resilience
Bryan Briney | Scripps Research
De novo discovery of fully human monoclonal antibodies
Abstracts
Guiding generative AI with molecular dynamics to design molecules with real activity
Rose Yu | UC San Diego
Generative AI can design molecules faster than we can evaluate them. Unfortunately most of the AI designed molecules don't show real activity. In this talk, I will showcase our work guiding generative AI models using molecular dynamics simulations and chemistry tools, which leads to strong compounds with desirable properties.
AI on "metastasis-in-a-dish" organoid screens to find new cancer therapies
Sural Ranamukhaarachchi | Serafin Labs
At Serafin, we're developing novel therapies to eradicate metastatic cancer, which accounts for more than 90% of cancer-related deaths. We discover these therapies using Osevia, a "metastasis-in-a-dish" 3D screening platform that reproduces complex metastatic behaviors not captured by traditional drug screening tools. To make these rich 3D tumors measurable at scale, we are building AI infrastructure that applies modern computer vision to extract and quantify features from metastatic cancer organoids, then integrates these phenotypic measurements with perturbation and molecular data to inform drug design. In this talk, I'll share how AI enables us to uncover new biological insights embedded in complex cellular phenotypes and translate these into actionable decisions for therapeutic development.
Reading DNA methylation across human tissues
Manoj Hariharan | Salk Institute
Two case studies illustrate the potential of AI/ML approaches applied to DNA methylation datasets for extracting biological insight. In the first, we generated single-nucleus data from 171 PBMC samples across 110 individuals, revealing distinct epigenetic signatures associated with various viral and bacterial infections and environmental exposures; these predictive biomarkers based on just a few genomic loci, led to the development of field-deployable assays. In the second, we examined DNA methylation dynamics in blood and skeletal muscle across acute exercise, training adaptation, and detraining, uncovering distinct methylation patterns stratified by intensity and sex that correlated with cell-type composition shifts and other molecular phenotypes. Together, these multi-parameter measurements highlight the potential for personalized training dosage prescriptions.
Your Body Is a Time Series: AI, Circadian Rhythms, and Precision Health
Emily Manoogian | Salk Institute
Every day, the body generates continuous data through sleep, activity, food intake, light exposure, and glucose fluctuations. In this talk, I will discuss how AI can integrate data from smartphones, wearables, continuous glucose monitors, and metabolic tests to uncover how circadian disruption—including shift work and jet lag—affects health. Drawing from our human studies, I will highlight how AI could help predict individual vulnerability and guide personalized interventions involving the timing of eating, exercise, light, and sleep. Ultimately, treating the body as a dynamic time series could transform circadian science into precision health.
Wiring AI into a Structural Biology Lab: Research, Operations, and Resilience
Charlie Bowman | Ward Lab, Scripps Research
Modern AI models are increasingly capable, but turning any new technology into dependable lab infrastructure is an engineering problem, not a plug-and-play one. In this talk, I'll describe how the Ward Lab is adopting AI across three parts of our work, and the technical challenges we've hit along the way. First, I'll share successes of our researchers using Claude Science to shorten the path from raw data to interpretable results by automating some of the tedium of structural biology. Then, a focus on operations, where modern AI coding models let a small team design, build, and maintain the specialized software and infrastructure that a lab our size could not otherwise staff for or sustain. Finally, on the topic of resilience I will share experiences with open-weight models, retrieval systems, and locally run agentic workflows to harden our pipelines against a fast-shifting model landscape, changing APIs, and data-governance constraints. Throughout, I'll focus on what actually worked, what broke, and the engineering decisions in between, sharing a practical look at running AI as real lab infrastructure.
De novo discovery of fully human monoclonal antibodies
Bryan Briney | Scripps Research
While generative binder design models are quite capable of generating antibody-like molecules against targets of interest, synthetic binders have historically faced major developability hurdles that prevent them from becoming effective drugs. Fully human antibodies are much more promising therapeutic candidates and can also be used as templates to guide rational vaccine design. We have developed an innovative structure-based in silico discovery pipeline, which we have used to identify experimentally validated binders against a variety of targets.
Sponsored by LatchBio. A San-Francisco based biotechnology company building agents and benchmarks for biological data analysis. Learn more at www.benchmarks.bio and https://latch.bio/jobs.