GenBio AI

I am glad to give a talk on "A World Model of the Virtual Cell" this Friday (May 29th) at the 2026 Cold Spring Harbor Laboratory Symposium. The outlook of an AI-driven Digital Organism (AIDO), such as a virtual cell, has recently captivated much excitement and imagination from both AI and Biology communities.   But what constitutes a meaningful realization of virtual cell? I will present an operational definition of the virtual cell based on World Model — a modern architecture recently emerged in AI research that supports advanced capabilities such as action-conditioned simulation, counterfactual reasoning, and long-horizon planning in complex dynamic environments. When applied to biological scenarios, a world model of the virtual cell is a generative model that simulates biological possibilities of a cell under any natural or artificial interventions, or a cell population (within a tissue type or an organ). A virtual cell world model (VCWM) contrasts predictive foundation models on specific tasks, such as gene-expression perturbation prediction, as seen in some recent definitions of the virtual cell. I will present a novel architecture for such a world model that enables simulated cell as an end-to-end platform: from actionable biological prompts to anticipated outcomes at all levels —molecular, structural, interactional, and morphological, in a fully aligned, integrative, multimodal, and multi-scale fashion, and prevail our first implementation of this system. A manuscript on VCMW can be found here: https://lnkd.in/dXdwyZsf Peter Koo, Le Song, Ziv Bar-Joseph, Emma Lundberg

We're #hiring a new Vice President of Engineering in Palo Alto, California. Apply today or share this post with your network.

We're #hiring a new Accounting Manager in Palo Alto, California. Apply today or share this post with your network.

Virtual cells are supposed to help drug discovery. Why aren't they evaluated on drug discovery tasks? In our new preprint "Cell-Level Virtual Screening," we investigate this and other fundamental questions about practical applications of virtual cells for drug discovery: 1. If we let virtual cells prioritize therapeutic candidates, would they do better than traditional molecular screens? 2. Is gene expression even a good representation of drug effects? To answer this, we curate two new benchmarks allowing us to fairly compare molecular and cell-level screening methods. We identify shortcomings in both types of methods, motivating us to go beyond static expression snapshots. We develop CellVS-Net, an amortized network estimator that predicts how gene networks restructure in response to unseen perturbations, substantially improving hit prioritization over molecular and expression baselines. Pre-print: https://lnkd.in/gYYrYw53 Congrats to my co-authors Sohan Addagudi, Jiaqi Wang, Ben Lengerich, Eric Xing Last year when we published our work using sample-specific gene networks to improve prognostic tumor subtyping in PNAS, we wondered if contextualized networks could also help prioritize treatments. This is an exciting step forward using personalized models to make precision treatment recommendations. Last year's article: https://lnkd.in/eHV28-Ny

We're #hiring a new Research Scientist Intern in Palo Alto, California. Apply today or share this post with your network.

✅ 𝐈𝐄𝐂 𝐒𝐭𝐚𝐫𝐭𝐮𝐩 𝐔𝐩𝐝𝐚𝐭𝐞𝐬 GenBio AI makes a feature in #MBZUAI President Professor Eric Xing’s interview with The Washington Post, highlighting the future of AI in medicine. GenBio AI is pioneering the concept of the 𝐯𝐢𝐫𝐭𝐮𝐚𝐥 𝐜𝐞𝐥𝐥, a world model of biology capable of predicting, simulating, and programming cellular processes at multiple levels. Through digital disease modeling and computational evaluation of treatments, this would allow scientists to explore and test ideas before real-world application. In the coming weeks, they will be releasing a 𝐯𝐢𝐫𝐭𝐮𝐚𝐥 𝐭𝐫𝐢𝐚𝐥 𝐩𝐫𝐨𝐭𝐨𝐭𝐲𝐩𝐞 that enables researchers to simulate the effects of designed drugs–alongside experimental conditions such as gene edits, small molecules, and antibody therapies–on cell behavior. Their long-term objective is to transform medical research into a high-throughput, automated endeavor, compressing research timelines that used to take decades. Read the full article here: https://lnkd.in/esGJpCWw #AI #ArtificialIntelligence #Entrepreneurship #IEC

Eric Xing has been named a 2026 fellow of the ISCB - International Society for Computational Biology, recognizing his contributions to computational biology, artificial intelligence and the development of data-driven approaches to complex biological systems. The ISCB Fellows program honors individuals who have demonstrated excellence through leadership, research and service, and whose work has advanced the field of computational biology. Fellows are selected for their sustained impact on scientific innovation and collaboration across disciplines. A professor in the Machine Learning Department, Computer Science Department and Language Technologies Institute, Xing develops machine learning and statistical methodologies and large-scale computational systems to address challenges in automated learning, reasoning and decision-making. His work spans applications in genomics, healthcare and artificial intelligence, with a focus on integrating complex biological data into predictive and interpretable models. He has advanced approaches that combine biological foundation models with AI-driven frameworks to analyze genomics, transcriptomics and clinical data, enabling more precise disease understanding and simulation-based exploration of complex biological systems. Xing will be formally recognized as part of the ISCB Fellows Class of 2026 at the ISMB 2026 conference this summer in Washington, D.C. For more information, visit the ISCB website.

How to transform drug design and treatment of diseases into a high-throughput and automated inference? — by digitally simulate disease and then run by different cures and drugs in a computational fashion using a virtual cell world model. This is GenBio AI ‘s mission for future medicine. Stay tuned for an imminent release. Forget about sequence to structure prediction or reverse folding, forget about perturb prediction of RNA counts, simulate a whole cell virtually, from molecular to morphological details and sdynamics. Le Song, Ziv Bar-Joseph, Emma Lundberg

We're #hiring a new Research Scientist Intern in Paris, Île-de-France. Apply today or share this post with your network.