Fwd: Mark Gerstein's talk at the CRG (July 21st 2026)

TITLE: Topics in Neurogenomics: Using AI to Study Endophenotypes & then Having it Take Over ABSTRACT: This talk explores the use of AI to study and eventually drive neurogenomics research. It covers linking wearable-derived "digital phenotypes" to genotype for conditions like ADHD, showing this approach uncovers SNPs missed by traditional case-control GWAS. It then examines single-cell data (e.g., PsychENCODE's 388-brain dataset) to build cell-type-specific regulatory and cell-to-cell communication networks, integrating them into deep learning models that predict disease from genotype and suggest drug targets. Finally, it turns to having AI do the science itself — automatically generating bioinformatics code, coordinating multi-agent LLM systems for unified reasoning, and considering the future risks of AI scientists in biomedicine. eur26+CRG

Abstract for my talk at UB (eur26+UB)

TITLE: The Changing Challenges in Human Genome Analysis ABSTRACT: This talk covers four areas: the EN-TEx resource for haplotype-aware genome analysis, classical genome annotation (pseudogene epigenetics across tissues), linking genetic variants to function via an allele-specific event catalog and a predictive transformer model, and new challenges in genomics -- privacy-preserving methods for handling population-scale disease genome data (homomorphic encryption, private federated learning, and leakage measurement). REFS: https://papers.gersteinlab.org/papers/NoisyFlow https://papers.gersteinlab.org/papers/HEPRS https://papers.gersteinlab.org/papers/epiPgene https://papers.gersteinlab.org/papers/Entex