📢 Bio-Babel: autonomous cross-language reconstruction of agent-ready computational biology ecosystems. Scientific software is often locked in its source language, callable from others but not natively extensible. Bio-Babel rebuilds it natively in the target ecosystem and makes it agent-callable, shipping each package with an agent-readable contract of its usage. Reconstructing a hierarchical, interdependent R-to-Python single-cell stack, Bio-Babel reproduced the originals faithfully and revealed pancreatic differentiation defects under graded SWI/SNF loss. Led by Nianping Liu and Xuanzhi Chen, with Miao Cui, Xiaoying Liao, Xiaoke Song, Weize Xu and Xiaojie.
📚 New review: The Landscape of Single-Cell Foundation Models: Design Principles, Applications, and Open Challenges, led by Jiayuan Ding. A broad community effort surveying how single-cell foundation models are designed, where they are being applied, and what still stands in the way, with Peter Pao-Huang, Yifan Lu, Weize Xu, Mengchen Wang and Xiaojie among the authors.
🎉 The Virtual Embryo Challenge: Generative Modeling of Embryogenesis Across Space, Scale and Time has been accepted to the NeurIPS 2026 Competition Track, one of 16 competitions selected. Led by Jiayuan Ding and Yifan Lu, the challenge is the first large-scale benchmark for predictive embryogenesis modeling, built on a multimodal mouse embryo dataset of roughly one million cells across 11 developmental stages. It also includes an Agent Team track, in which autonomous coding agents iteratively improve their own predictive models.
📄 Predictive single cell foundation model for gene regulation and aging with privacy-preserving tabular learning, now under journal submission. Tabula pairs tabular modeling with federated learning, and Chiron lets institutions train together without sharing raw data. Applied to paired young and aged human fibroblasts, Tabula nominates rejuvenation factors that conventional approaches miss. Led by Jiayuan Ding, with colleagues including Yuancheng Ryan Lu.
📢 Generative Modeling of Mouse Embryogenesis for Fate and Disease Prediction. We introduce Navigo, which learns a developmental vector field by combining population-level flow matching with molecular RNA kinetics, across a mouse atlas of 43 time points and 12.4 million cells. Navigo resolves regulatory networks distinguishing congenital heart disease subtypes and predicts perturbation effects zero-shot, validated on six knockout genotypes. Led by Yimin Fan, with Zehua Zeng, Lei Li and Xiaojie.
📄 Generative Modeling with Flux Matching, by Peter Pao-Huang with Xiaojie and Stefano Ermon. Flux Matching generalizes score-based models to vector fields that need not be conservative, turning the vector field itself into a design choice rather than a fixed target, and unlocking faster sampling and more interpretable, mechanistic models.
🎉 Tabula, our tabular self-supervised foundation model for single-cell transcriptomics, has been published in Advances in Neural Information Processing Systems. Congratulations to Jiayuan Ding and the team! Tabula reaches state-of-the-art performance on gene imputation, perturbation prediction, cell type annotation and multi-omics integration, using only half the pretraining data.
Xiaojie has been selected as a recipient of the Laude Institute Moonshot program. Together with Stanford Profs. Emily Fox and James Zou, and Harvard Prof. Marinka Zitnik, we will pioneer The Virtual Embryo Moonshot: Predictive Modeling of Human Development and Congenital Defects.
📄 Towards predictive virtual embryos with genomics and AI is out in Nature Methods, by Natalie Cao, Yifan Lu and Xiaojie. The piece lays out how predictive virtual embryo systems that combine single-cell and spatial data with AI could let us model mammalian embryogenesis across scales.
📢 PantheonOS: an evolvable multi-agent framework for automatic genomics discovery. PantheonOS is an evolvable, privacy-preserving multi-agent framework that reconciles generality with domain specificity, and its agentic code evolution reaches super-human performance on batch correction and gene panel selection. Led by Weize Xu, with Erwin Poussi, Zehua Zeng, Christopher Zou, Yifan Lu, Miao Cui, Cinlong Huang, Jiayuan Ding, Naoki Konno, Nianping Liu, Lei Li and many collaborators.
🎉 Xiaojie has been awarded a Wu Tsai Neurosciences Institute Big Ideas in Neuroscience grant, together with Stanford Profs. Thomas Clandinin and Will Allen, for In Vivo Functional Genomics and Foundational Virtual Models of Brain Homeostasis and Resilience. The project uses cross-species genomics and AI to understand how brain cells stay resilient to stress and why that capacity declines with age, and is co-sponsored by the Knight Initiative for Brain Resilience.
Huge congratulations to Naoki Konno, who has been awarded the 2026 AHA Postdoctoral Fellowship for his work Deciphering Congenital Heart Defects Through Single-Cell Evolution and in silico Perturbation! 🎉
📢 MERFISH+, a large-scale, multi-omics spatial technology resolves the molecular holograms of the 3D human developing heart. A large-scale spatial multi-omics technology applied to the 3D human developing heart, with Yifan Lu, Zehua Zeng and Xiaojie among the authors.
We have received a California Institute for Regenerative Medicine DISC0 grant, awarded to Xiaojie together with BASE Affiliate Faculty Member Dr. Matteo Molè, for Dissecting the cellular and molecular interactions between embryo and endometrium during human implantation. Read more about the CIRM Discovery Research awards.
Xiaojie and BASE Affiliate Faculty Member Dr. Matteo Molè have been selected for the ISCBRM Stinehart-Reed Collaborative Seed Grant Program, supporting our joint work on A Spatial Transcriptomic Atlas of Embryo-Endometrial Crosstalk During Human Implantation. The project pairs an advanced model of the human uterus with high-resolution 3D gene mapping to build the first detailed picture of embryo-mother interactions during early pregnancy, with the aim of improving IVF outcomes.
We are really excited to share that Peter Huang’s rotation project, Expressive Geometric Generative Modeling with Noise-Conditioned Graph Networks, has just been accepted as a poster presentation at ICML! This marks Peter’s first publication as a PhD student, and we are incredibly proud of his initiative, creativity, and perseverance in making this achievement possible. We can’t wait to see what he’ll do next!
Welcome to four incredibly talented visiting researchers who have recently joined our lab: Chen Li from Dr. Xuegong Zhang’s lab at Tsinghua University, Qiangwei Peng from Dr. Tiejun Li’s lab at Peking University, Valentina Giunchiglia from Dr. Adam Hampshire’s lab at Imperial College London and Dr. Marinka Zitnik’s lab at Harvard University, and finally, Zehua Zeng, developer of the impressive OmicsVerse framework, from Dr. Hongwu Du’s lab at the University of Science and Technology Beijing.
Chen will be working on novel approaches to perturbation prediction using foundation models. Qiangwei will focus on modeling mechanical forces during embryogenesis. Valentina will contribute to developing the next generation of multi-modal foundation models. Zehua will dedicate his efforts to building robust frameworks for spatial multi-omics integration.
📢 Paper Alert!
We’re thrilled to share our lab’s first single-cell foundation model: Tabula, developed by the incredible Jiayuan Ding!
Tabula introduces a novel pretraining strategy, tabula learning, and integrates federated learning to achieve state-of-the-art performance. Remarkably, it outperforms all existing foundation models using only half the training data. It also demonstrates strong predictive power in uncovering both pairwise and higher-order combinatorial gene regulatory patterns.
We are thrilled to share that our first paper from the lab, Spateo for spatiotemporal modeling of molecular holograms, is now online in Cell. Spateo is a comprehensive analytical framework for 3D whole-embryo spatiotemporal modeling. This work is also highlighted in Nature and Nature Methods. Huge congratulations to Yifan for his phenomenal contribution to this work, from building the state-of-the-art 3D reconstruction to pioneering a powerful morphogenesis vector field approach. This wouldn’t have been possible without your brilliance and dedication! 🎉
Xiaojie has been selected for this year’s NIH Director’s New Innovator Award. We are pioneering new approaches that combine machine learning, single-cell/spatial genomics, and systems biology to create predictive models of the virtual heart and whole embryo. See the Stanford news story.
We’re excited to welcome Yuexuan Yang (Bioengineering) and Peter Huang (Computer Science) as rotation students! Yuexuan will work on optics-free spatial transcriptomics while Peter will focus on diffusion models for spatiotemporal dynamics.
We are thrilled to welcome two talented undergraduate students to our team: Dingcheng Yi from Peking University and Mengchen Wang from Tsinghua University. Dingcheng has an outstanding academic record, ranking No. 2 in his program. Mengchen is part of Tsinghua’s prestigious Yao Class, a special pilot program under the tutelage of Turing Award winner Andrew Chi-Chih Yao. We are excited to have Dingcheng and Mengchen join us in exploring 3D spatial transcriptomics from different perspectives!
We are thrilled to welcome two talented students join us today: Yifan Lu, a visiting PhD student from Wuhan University in China; Stephen Zhang, a visiting PhD student from the University of Melbourne in Australia. Yifan will be exploring the exciting frontier of applying computer vision approaches to 3D spatial transcriptomics. Stephen has made some exceptional work in single-cell dynamics modeling using optimal transport and is planning to develop spatiotemporal models of single cells at the whole organ or embryo level.
Exciting news! We have received a CIRM grant, and presumably ranked No. 1 based on the prior year’s criteria! We will develop single cell foundation model to understand endothelial cell fate decisions for pulmonary arterial hypertension.
Eric Waters, the new lab manager for the three BASE labs (Gifford, Engreitz, and Qiu), has just started his adventure with us. With a rich background spanning Human Physiology to Computer Science and data science, he is eager to help set up our wet lab space and support the many exciting research projects in our lab. Welcome aboard, Eric!
Cinlong Huang, a recent graduate from UC Berkeley, has started working in the Qiu lab as a research assistant. Cinlong will help us set up the wet lab space and explore exciting new frontiers in spatially and temporally resolved single-cell multi-omics. We are delighted to have Cinlong join us for our exciting journey ahead. Welcome, Cinlong!
Samuel King from Bioengineering department at Stanford started his rotation in the lab. Samuel is excited about the opportunity to work on synthesizing single cell / spatial genomics with machine learning approach to decipher regulatory mechanism of embryogenesis and organogenesis. We are excited to have Samuel to join the team as the first rotation student. Welcome, Samuel!
Mengchen Wang from Tsinghua University, China, started a brief stint with us from December 23, 2023, to January 2, 2024. Mengchen is keenly interested in exploring new frontiers of transformer model for single-cell and spatial genomics. He will be rejoining us for a more extended visit in the summer. Welcome, Mengchen!
Yuzhen Mao from Simon Fraser University in Canada joined me on my first day at Stanford as a visiting Master’s student until the end of March 2024. Yuzhen will pursue his passion in foundational models of single cells. Welcome, Yuzhen!
The Qiu Lab officially started at Stanford on December 16th, 2023. We are recruiting talents at all levels to join our dynamic team. Explore our current lab openings and become a part of our exciting journey!