Xiangjie Zhao UNC Chapel Hill

Research program

From perturbation to mechanism.

Technology development, cardiovascular biology, and predictive computation as one loop: build causal measurements, discover regulatory mechanisms, then predict the next intervention worth running.

The experimental cycle Five stages — question, perturbation, multi-omic readout, model, validation — connected in a loop that feeds new hypotheses back to the start. new hypotheses 010203 0405 Question Perturbation Multi-omic readout Model Validation what is causal? CRISPR · AAV RNA · ATAC · spatial predict response in vivo test
The loopBuild causal measurements, discover regulatory mechanisms, predict the next intervention — then test it.
01

Perturbation Omics Technologies

Make cellular causality measurable.

Experimental platforms that pair programmed perturbations with transcriptomic, epigenomic, and spatial phenotypes.

Perturb-seqPerturb-multiomePerturb-spatial
Research rationale

Single-cell atlases reveal biological heterogeneity but are primarily observational. Perturbation omics moves from association to mechanism by measuring controlled interventions across thousands of individual cells.

The long-term goal is a unified platform that progresses from pooled discovery to multimodal mechanism and spatial validation.

02

Cardiovascular Functional Genomics

Resolve the regulatory logic of the heart.

Causal maps of the programs that govern cardiac maturation, maladaptive remodeling, and regenerative potential.

Cardiac maturationHeart failureRegeneration
Research rationale

Cardiovascular development and disease involve coordinated changes across cardiomyocytes, fibroblasts, vascular cells, immune cells, and other populations.

My future program will connect human genetics, physiological models, multi-omic measurements, and perturbation screens to experimentally testable mechanisms.

03

AI-enabled Biology

Let experiments train the model.

Interpretable computational systems that predict cellular responses and guide more informative experiments.

Perturbation predictionComputational frameworksAI agents for biology
Research rationale

Perturbation datasets contain defined interventions and therefore provide an unusually strong foundation for biological prediction.

Models will be evaluated by whether they identify mechanisms, generalize across contexts, and prioritize experiments that produce new knowledge.

Experiment–model feedback Measured perturbation responses train a model; the model predicts unseen responses, which decide the next experiment. Measured perturbation × cell Model interpretable, testable Predicted unseen perturbations design the next experiment
03Conceptual framework · experiment–model feedback

Current projects

Work in progress.

Descriptions stay deliberately brief while studies are unpublished.

Under wraps

Perturbation studies in progress

Several are running. Designs and results stay off this page until the corresponding papers are published.

In preparation

Cardiac maturation snATAC atlas

Chromatin dynamics across postnatal cardiac maturation.

Collaboration

Interested in perturbation genomics, cardiovascular models, or predictive biology?

Start a conversation See the resources