Single-cell and spatial genomics, meet machine learning

A multidisciplinary lab at Stanford University, modeling heart evolution, development and disease across space, scale and time.

Mission of the Qiu Lab

We work to unravel and predict the gene regulatory networks and cell-cell interactions that drive mammalian cell-fate transitions over time and space, with a special emphasis on heart evolution, development and disease. By pairing the interpretability of systems biology with the scalability of machine learning, we aim to demystify developmental biology for human health and medicine.

Computation

Transformers, neural ODEs and diffusion models, paired with interpretable systems-biology methods.

Experimentation

Metabolic-labeling scRNA-seq, ultra-high-resolution spatial transcriptomics, Perturb-seq and lineage tracing.

Prediction

Differentiable models that forecast cell states and fate trajectories under genetic and environmental perturbation.

The Qiu Lab pioneered in developing several powerful computational software, such as Monocle 2/3, Dynamo, and Spateo for the emergent single-cell and spatial genomics datasets, and is now building a new generation of AI-native platforms, including PantheonOS and Bio-Babel. See more below:

Tools & Platforms

The Qiu Lab builds open-source software for single-cell and spatial genomics, from the analysis packages that helped establish the field's methods to a new generation of AI-native platforms.

AI-native platforms

PantheonOS the AgentOS for biological discovery
- An evolvable, privacy-preserving multi-agent framework for automatic genomics discovery.
- LLM-powered agents collaborate on specialized scientific tasks, reconciling general-purpose reasoning with deep domain specificity.
- A four-layer architecture: application, interface, agent, LLM.
- The Pantheon Store offers 2,000+ curated biomedical agents, teams and skills.
Bio-Babel the classics of bioinformatics, in more than one tongue
- Rebuilds classic tools natively in a new language rather than bridging to them at runtime, so they can be extended and composed, not merely called.
- Every package ships a machine-readable contract, so an agent that has never seen it still calls it correctly.
- 17 libraries live on PyPI, 1,606 contracted symbols, zero runtime bridges.

Analysis packages

Spateo
- Multiscale modeling of spatial genomics, from subcellular to whole embryo.
Dynamo
- Predicts cell-fate transitions, optimal reprogramming factors and in silico perturbations.
Dynast
- Efficient quantification for metabolic-labeling scRNA-seq.
Monocle 2/3
- Infers developmental trajectories from single-cell data.
Scribe
- Learns causal gene-regulatory networks.
Aristotle
- The end-to-end ecosystem tying these together for quantitative spatiotemporal modeling.

Monocle was developed during Dr. Qiu's Ph.D. with Dr. Cole Trapnell at the University of Washington, Scribe with Sreeram Kannan and Arman Rahimzamani at UW, Dynamo as a postdoctoral fellow with Dr. Jonathan Weissman at the Whitehead Institute and MIT together with Yan Zhang from Dr. Jianhua Xing's lab at Pitt, and Dynast with Joseph Min.

Qiu Lab @ Stanford

Virtual Embryo Challenge (Open Now)