John Wu

Hello! I’m John Wu, a Ph.D. student in the Computer Science Department at the University of Illinois Urbana-Champaign. I’m researching AI applications in healthcare under the guidance of Professor Jimeng Sun.

Prior to my Ph.D., I worked as an undergraduate researcher in Dr. Das’ mathematical immunology lab, focusing on parameter estimation methods for mechanistic modeling. My work centered on two key aspects: handling noisy data regimes using the generalized method of moments and accelerating parameter estimation through parallelization and surrogate modeling with deep neural networks. Although deep learning methods often surpass mechanistic models in computational efficiency, they lack human interpretability and well-defined internal behavior. Consequently, deploying them in high-stakes applications like healthcare requires extensive testing and supervision, which can be prohibitively expensive.

Currently, I’m deeply interested in mechanistic interpretability, a rapidly growing field that bridges the gap between computational complexities and comprehensive model explanations. Recognizing the nontrivial consequences in healthcare and the importance of patient trust, my goal is to develop useful models that are more explainable.

selected publications

BioNetGMMFit: estimating parameters of a BioNetGen model from time-stamped snapshots of single cells

John Wu, William C. L. Stewart, Ciriyam Jayaprakash, and 1 more author

npj Systems Biology and Applications, Sep 2023

Abs HTML PDF

Mechanistic models are commonly employed to describe signaling and gene regulatory kinetics in single cells and cell populations. Recent advances in single-cell technologies have produced multidimensional datasets where snapshots of copy numbers (or abundances) of a large number of proteins and mRNA are measured across time in single cells. The availability of such datasets presents an attractive scenario where mechanistic models are validated against experiments, and estimated model parameters enable quantitative predictions of signaling or gene regulatory kinetics. To empower the systems biology community to easily estimate parameters accurately from multidimensional single-cell data, we have merged a widely used rule-based modeling software package BioNetGen, which provides a user-friendly way to code for mechanistic models describing biochemical reactions, and the recently introduced CyGMM, that uses cell-to-cell differences to improve parameter estimation for such networks, into a single software package: BioNetGMMFit. BioNetGMMFit provides parameter estimates of the model, supplied by the user in the BioNetGen markup language (BNGL), which yield the best fit for the observed single-cell, time-stamped data of cellular components. Furthermore, for more precise estimates, our software generates confidence intervals around each model parameter. BioNetGMMFit is capable of fitting datasets of increasing cell population sizes for any mechanistic model specified in the BioNetGen markup language. By streamlining the process of developing mechanistic models for large single-cell datasets, BioNetGMMFit provides an easily-accessible modeling framework designed for scale and the broader biochemical signaling community.
DILA: Dictionary Label Attention for Mechanistic Interpretability in High-dimensional Multi-label Medical Coding Prediction

John Wu, David Wu, and Jimeng Sun

Sep 2024

HTML PDF
EMNLP

Beyond Label Attention: Transparency in Language Models for Automated Medical Coding via Dictionary Learning

John Wu, David Wu, and Jimeng Sun

In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing , Nov 2024

HTML