About

Hui Li, MD, MS, is a Research Associate Professor of Medicine in the Hematology-Oncology department at the Perelman School of Medicine, University of Pennsylvania. His research focuses on HIV/AIDS, specifically on the structural and genetic basis of HIV-1 envelope recognition by broadly neutralizing antibodies, as well as the coevolution of Env-antibodies and the development of HIV-1 vaccine strategies. He is involved in the Single Genome Sequencing and SHIV Design Core Service Center at the Penn Center for AIDS Research. Dr. Li's educational background includes an M.D. in Stomatology from Capital University of Medical Sciences in Beijing, China, obtained in 1992, and an M.S. in Computer Science from the University of Alabama at Birmingham, completed in 2003. His work has contributed to understanding HIV-1 neutralization mechanisms, B cell responses, and vaccine design, with multiple publications and presentations in the field of immunology and virology.

Research topics

• Gastroenterology
• Medicine
• Internal medicine
• Biochemistry
• Microbiology
• Biology
• Genetics
• Bioinformatics
• Chemistry

Selected publications

• Transfer Learning for Ridge Regression with Random Coefficients

  Statistica Sinica · 2026-04-16

  articleOpen accessSenior author

  Ridge regression with random coefficients provides a flexible approach for modeling many small but nonzero effects in high-dimensional data.We embed this framework in transfer learning by leveraging source samples from related regression models: the informativeness of each source is captured via the correlation between its coefficients and those of the target.We propose two weighted estimators-one minimizing estimation risk and the other minimizing prediction risk-each formed as an optimal blend of target and source ridge estimates.Under the high-dimensional regime p/n , where p is the number of the predictors and n is the sample size, random matrix theory yields closed-form limits for these optimal weights and their associated risks.Through simulations and applications to lipid-trait and colorectal-cancer microbiome prediction, our methods consistently outperform both target-only and pooled-data ridge regression.

  PublisherDOI

• Robust causal gene network estimation for large-scale single-cell perturbation screens using reduced control function

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-04-21

  articleSenior author

  Abstract Single-cell CRISPR perturbation screens offer a powerful framework for causal discovery in gene regulatory networks, but existing methods struggle with high-dimensional count data, unmeasured confounding, and the increasing prevalence of high-multiplicity-of-infection (MOI) designs. We introduce RICE, a scalable framework for causal gene network estimation that integrates a reduced control function to address latent confounding with a constrained generalized linear model accommodating both hard and soft interventions. By enforcing differentiable acyclicity constraints, RICE enables efficient GPU-based optimization for large-scale data. Across synthetic benchmarks, RICE achieves higher accuracy and robustness than existing methods and remains stable under strong confounding and high-MOI settings. Applied to multiple single-cell perturbation datasets, including CRISPRi screens in K562 and RPE1 cells and a Perturb-CITE-seq data set with CRISPR-Cas9 knockout (KO), RICE recovers biologically coherent networks with edge weights consistent with perturbation effects and enriched for known regulatory interactions. These results establish RICE as a flexible and scalable approach for causal discovery in modern single-cell perturbation studies.

  PublisherDOI

• Test of partial effects for Frechet regression on Bures-Wasserstein manifolds

  ArXiv.org · 2025-06-30

  preprintOpen accessSenior author

  We propose a novel test for assessing partial effects in Frechet regression on Bures Wasserstein manifolds. Our approach employs a sample splitting strategy: the first subsample is used to fit the Frechet regression model, yielding estimates of the covariance matrices and their associated optimal transport maps, while the second subsample is used to construct the test statistic. We prove that this statistic converges in distribution to a weighted mixture of chi squared components, where the weights correspond to the eigenvalues of an integral operator defined by an appropriate RKHS kernel. We establish that our procedure achieves the nominal asymptotic size and demonstrate that its worst-case power converges uniformly to one. Through extensive simulations and a real data application, we illustrate the test's finite-sample accuracy and practical utility.

  PublisherOA PDFDOI

• Relationship Between Timing of Hemodialysis and Procedure-Related Complications Following Endoscopy in Patients with End-Stage Renal Disease

  Digestive Diseases and Sciences · 2025-06-18 · 2 citations

  articleOpen access

  PURPOSE: There is a growing population of patients with end-stage renal disease (ESRD) on hemodialysis (HD) requiring endoscopic procedures. While literature supports timing general surgeries one day following dialysis, there is inadequate data on optimal timing of endoscopic procedures. This study examined the procedural complications related to endoscopic procedures based on timing post-HD. METHODS: This retrospective cohort study included endoscopic procedures in patients with ESRD on thrice-weekly HD at a single institution. Procedural complications were analyzed using a GEE model incorporating inverse probability treatment weighting logistic regression based on pre-procedural covariates (Model 1) and combined pre- and intra-procedural covariates (Model 2). RESULTS: 252 procedures were identified in 191 unique patients, among which 48, 147 and 57 were performed 0, 1 and 2 days post-HD, respectively. Procedures performed on the same day post-HD were more likely to be inpatient (n = 44, 91.7% vs. n = 101, 68.7% vs. n = 39, 68.4% for 0 vs. 1 vs. 2 days post-HD, respectively; p = 0.005) with indication for GI bleeding and/or anemia (n = 40, 83.3% vs. n = 95, 64.6% vs. n = 42, 73.7% for 0 vs. 1 vs. 2 days post-HD, respectively; p = 0.04). Patients undergoing procedures 0 compared to 1-2 days post-HD were more likely to experience mortality (Model 1 OR 2.91, 95% CI 1.24-6.80; p = 0.01; Model 2 OR 3.22, 95% CI 1.42-7.29, p = 0.005). CONCLUSION: Endoscopic procedures performed on the same day following HD may be associated with a higher risk for mortality. Further studies are needed to reproduce these findings and explore the underlying mechanisms driving this association.

  PublisherOA PDFDOI

• Multicalibration for modelling censored survival data with universal adaptability

  Biometrika · 2025-01-01 · 1 citations

  articleSenior author

  Summary Traditional statistical and machine learning methods typically assume that the training and test data follow the same distribution. However, this assumption is frequently violated in real-world applications, where the training data in the source domain may underrepresent specific subpopulations in the test data of the target domain. This article addresses target-independent learning under covariate shift, focusing on multicalibration for survival probability and restricted mean survival time. A black-box post-processing boosting algorithm specifically designed for censored survival data is introduced. By leveraging pseudo-observations, our method produces a multicalibrated predictor that is competitive with inverse propensity score weighting in predicting the survival outcome in an unlabelled target domain, ensuring, not only overall accuracy, but also fairness across diverse subpopulations. Our theoretical analysis of pseudo-observations builds upon the functional delta method and the $ p $-variational norm. The algorithm’s sample complexity, convergence properties and multicalibration guarantees for post-processed predictors are provided. The results establish a fundamental connection between multicalibration and universal adaptability, demonstrating that our calibrated function is comparable to, or outperforms, the inverse propensity score weighting estimator. Extensive numerical simulations and a real-world case study on cardiovascular disease risk prediction using two large prospective cohort studies validate the effectiveness of our approach.

  PublisherDOI

• Antibiotic exposure is associated with minimal gut microbiome perturbations in healthy term infants

  Microbiome · 2025-01-24 · 10 citations

  articleOpen access

  BACKGROUND: The evolving infant gut microbiome influences host immune development and later health outcomes. Early antibiotic exposure could impact microbiome development and contribute to poor outcomes. Here, we use a prospective longitudinal birth cohort of n = 323 healthy term African American children to determine the association between antibiotic exposure and the gut microbiome through shotgun metagenomics sequencing as well as bile acid profiles through liquid chromatography-mass spectrometry. RESULTS: Stool samples were collected at ages 4, 12, and 24 months for antibiotic-exposed (n = 170) and unexposed (n = 153) participants. A short-term substudy (n = 39) collected stool samples at first exposure, and over 3 weeks following antibiotics initiation. Antibiotic exposure (predominantly amoxicillin) was associated with minimal microbiome differences, whereas all tested taxa were modified by breastfeeding. In the short-term substudy, we observed microbiome differences only in the first 2 weeks following antibiotics initiation, mainly a decrease in Bifidobacterium bifidum. The differences did not persist a month after antibiotic exposure. Four species were associated with infant age. Antibiotic exposure was not associated with an increase in antibiotic resistance gene abundance or with differences in microbiome-derived fecal bile acid composition. CONCLUSIONS: Short-term and long-term gut microbiome perturbations by antibiotic exposure were detectable but substantially smaller than those associated with breastfeeding and infant age.

  PublisherOA PDFDOI

• Lung transplant for CF: Low lung bacterial burden and immune mediators in year one associate with CLAD development

  Journal of Cystic Fibrosis · 2025-10-10

  article
  PublisherDOI

• P0173 The impact of dietary composition on metallothionein gene expression in the intestinal epithelium by exclusive enteral nutrition

  Journal of Crohn s and Colitis · 2025-01-01

  articleOpen access

  Abstract Background The mechanism by which exclusive enteral nutrition (EEN) is effective in the treatment of Crohn’s disease remains elusive. Epidemiologic studies have shown that a substantial proportion of patients with Crohn’s disease are deficient in micronutrients such as zinc, which has been shown to play an important role in the integrity of intestinal epithelial barrier function. Zinc plays a critical role in the transcriptional regulation of metallothioneins (MT), cysteine rich binding proteins that regulate both epithelial and immune function where they are generally believed to play a beneficial role in reducing disease. Methods Transcriptomic analyses were performed by both bulk RNAseq as well as GeoMx spatial transcriptomics. Enzymatic food digests using both bile acids and pancreatic enzymes, analyzed by Proton NMR metabolomics, were used to generate cell culture media for studies in intestinal epithelial cells. Results RNAseq of rectal biopsies obtained from healthy human subjects consuming EEN (Modulen-IBD, Nestlé Health Science) vs. an omnivore diet revealed a very specific induction of genes involved in mineral transport, namely MTs such as MT1G and MT2A. Spatial transcriptomics revealed that the induction of MTs was specific to the colonic epithelium. To develop a cellular model to characterize the mechanism underlying this response, we investigated the impact of animal protein, vegetable, and EEN based foods on the transcriptome of intestinal epithelial cells in culture by both RNAseq and targeted quantitative RT-PCR. This model involved the development of a physiologically relevant modality to digest whole food to generate serum-free cell culture media that was characterized using NMR-based metabolomics. Animal protein and EEN diets induced the expression of specific MTs whereas vegetable diets reduced their expression. We also show that zinc induced MT gene expression in intestinal epithelial cells whereas phytic acid, a phytochemical with known anti-nutrient properties that chelates zinc, reduces the ability of protein-based foods and EEN to induce MT gene expression. Conclusion The consumption of EEN leads to a very specific transcriptomic signature of MT gene expression in the rectal epithelium of humans that can be reproduced in cell culture by exposure to both animal protein-based foods and EEN. By contrast, vegetable-based foods inhibit MT gene expression possibly through the chelation of zinc. These results are consistent with the notion that the exclusion of plant-based foods through the consumption of EEN may have beneficial effects in the treatment of Crohn’s disease by reducing exposure to anti-nutrient phytochemicals that reduce the absorption of beneficial micronutrients such as zinc.

  PublisherOA PDFDOI

• Increased reflux secondary bile acids are associated with changes to the microbiome and transcriptome in Barrett’s esophagus

  Gut Microbes · 2025-08-22 · 8 citations

  articleOpen access

  showed the most associations with gene expression, including the oxidative phosphorylation pathway. We identified two distinct BE gene expression clusters independent of histology, bile acid, or microbiome composition. These findings suggest bile acids shape the BE microbiome and associate with gene expression changes potentially relevant to EAC development.

  PublisherOA PDFDOI

• Simultaneous Estimation of Many Sparse Networks via Hierarchical Poisson Log-Normal Model

  Journal of Computational and Graphical Statistics · 2025-10-10

  articleSenior authorCorresponding
  PublisherDOI

Frequent coauthors

• Dong Guo

  Hubei University of Technology

  49 shared

• Kyle Bittinger

  California University of Pennsylvania

  46 shared

• Tommaso Cai

  University of Oslo

  40 shared

• Meenakshi Bewtra

  University of Pennsylvania

  39 shared

• Gary D. Wu

  University of Pennsylvania

  37 shared

• Frederic D. Bushman

  35 shared

• James D. Lewis

  University of Pennsylvania

  34 shared

• Sumona Saha

  UW Health University Hospital

  34 shared

Education

• PhD, Statistics

  University of Washington

  1995