Jundong Li
University of Virginia, VA·Computer Science·Early-career
Much of the world's data is about connections: atoms bonded into molecules, people linked in social networks, records related to one another. Machine learning models that work on this kind of connected data are called graph models.
- Fields
Computer Science
Mathematics & Statistics
Biomedical Engineering
- Topics
graph machine learning
graph neural networks
graph transformers
- Grants
- 3 known grants, through Jul 2029
What they work on
- Design graph neural networks tailored to peptides, including amino acid-aware message passing and graph transformers, and combine them with pretrained protein language models to merge structural and evolutionary information.
- Build a large dataset of peptide 3D conformers using molecular dynamics sampling, and train geometry-aware networks that encode bond lengths, dihedral angles and backbone curvature.
- Create multimodal 2D–3D foundation models that can predict shape-dependent peptide properties even when only the 2D bond structure is available at prediction time.
- Predict and validate peptide chromatographic retention times, which matter for manufacturing, and antibacterial activity, and release the platform as open-source software.
- Develop post-hoc interpretation tools that diagnose whether graph learning models are fair, robust and causally sound, then use those explanations to guide training and improve model robustness and interactivity.
- Use model explanations to refine graph data, both by adding training examples through a counterfactual Mixup strategy and by compressing graphs through data distillation.
- Use relational knowledge about how data points relate to one another to relax unrealistic assumptions when estimating individual-level causal effects from observational data.
- Build causal inference models that incorporate meta knowledge about a dataset's properties and application knowledge about real-world needs, to improve personalized decision making.
More about the research
To be useful in science and in decisions about people, these models must be accurate, understandable, and able to tell cause from coincidence. The grants cover three linked directions. One applies graph machine learning to discovering peptides. One makes the explanations of graph models more useful. One uses human knowledge to improve causal inference, which estimates what effect a decision will have on a specific individual.
Why it matters
Faster computer screening of peptides could help researchers find candidate antimicrobial peptides and other biomaterials before spending time and money on lab experiments. Open-source tools and datasets would lower the barrier for other labs. Methods that check graph models for fairness and robustness, and that estimate how a decision affects a specific individual from observational records, could support better-informed decisions in many fields, including efforts to improve student success and retention.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
An Integrated 2D and 3D Graph Machine Learning Platform for Peptide Discovery
NSF · Aug 2026 – Jul 2029
Collaborative Research: III: Small: Graph-Oriented Usable Interpretation
NSF · Oct 2022 – Sep 2027
CAREER: Toward A Knowledge-Guided Framework for Personalized Decision Making
NSF · Sep 2022 – Aug 2027
Closest labs
Meng Jiang
University of Notre Dame, IN·Computer Science·Early-career
Develop and compare natural language generation models, one data-driven and one based on theories of abduction, for comparative summarization and argument generation in scientific domains.
Jurij Leskovec
Stanford University, CA·Computer Science
Develop multi-level graph neural networks that handle graphs whose nodes are themselves complex structures, such as molecular graphs, for scientific foundation models.
Philip Yu
University of Illinois Chicago, IL·Computer Science
Integrate open-source anomaly detection tools for tabular, graph and time-series data into a single OpenAD ecosystem with a standardized development environment for building and evaluating models.
The school
University of Virginia: 43 labs on PI Pages, $814M a year on research, 413 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.