Research
Developing medical AI methods, including AI agents and large language models (LLMs), to discover disease mechanisms
We develop medical AI methods, including AI agents and large language models, to integrate biomedical data and generate testable hypotheses about disease mechanisms. My laboratory has published multiple LLM-based algorithms for biological and medical applications at premier computer science conferences, including NeurIPS, ICML, AAAI, and ACL. We complement these computational approaches with comprehensive experimental validation and clinical investigation to translate AI-generated hypotheses into novel discoveries in disease biology and meaningful advances in clinical medicine.
Publication
Abdullah KG, et al. (2026) Gliomas phenocopy an inborn error of metabolism to drive neuronal activity and tumor growth. Cell S0092-8674(26)01010-X.
Lebek S, et al. (2023) Ablation of CaMKIIδ oxidation by CRISPR-Cas9 base editing as a therapy for cardiac disease. Science 379, 179–185.
Chai AC, et al. (2023) Base editing correction of hypertrophic cardiomyopathy in human cardiomyocytes and humanized mice. Nature Medicine 2023, 29, 401–411.
Huang L, et al. (2019) SR-B1 drives endothelial cell LDL transcytosis via DOCK4 to promote atherosclerosis. Nature. 569(7757):565-569.
Develop machine learning and deep learning algorithms to study single-cell and spatial multi-omics data.
We develop machine learning and deep learning algorithms to analyze and integrate single-cell and spatial multi-omics data, linking molecular variation within individual cells to the organization of tissues. Our methods aim to uncover cellular heterogeneity, gene regulatory programs, and interactions among cells that contribute to disease initiation and progression. By connecting these molecular and spatial patterns with clinical phenotypes, we generate testable hypotheses about disease mechanisms and identify candidate biomarkers and therapeutic targets for experimental and clinical validation.
Publication
Zhou Y, Xiao X, Dong L, Tang C, Xiao G#, Xu L# (2025) Cooperative integration of spatially resolved multi-omics data with COSMOS. Nature Communications 16(1):27.
Tang C, Zhou Y, Xiao X, Dong L, Yu L, Li Q, Xiao G#, Xu L#. (2025) 3D reconstruction of spatial transcriptomics with spatial pattern enhanced graph convolutional neural network. Briefings in Bioinformatics. 27(1):bbag060.
Guo Y, Zhu B, Tang C, Rong R, Ma Y, Xiao G#, Xu L#, Li Q# (2024) BayeSMART: Bayesian clustering of multi-sample spatially resolved transcriptomics data. Briefings in Bioinformatics 25(6):bbae524.
Jiang X, Wang S, Guo L, Zhu B, Wen Z, Jia L, Xu L#, Xiao G#, Li Q#. (2024) iIMPACT: integrating image and molecular profiles for spatial transcriptomics analysis. Genome Biology 25(1):147.
Develop integrated multi-omics analysis to identify new biomarkers and therapeutic targets in human disease.
My team develops integrated multi-omics approaches that combine genomic, epigenomic, transcriptomic, and proteomic data to uncover molecular mechanisms underlying human disease. By linking these molecular insights to clinical characteristics and outcomes, we aim to identify new biomarkers for diagnosis, prognosis, and treatment response, as well as therapeutic targets for experimental and clinical validation.
Publication
Sternisha AC, et al. (2026) A transcriptional biosensor reveals mechanisms of α-ketoglutarate signaling to chromatin. Science 393(6808):eadx8675.
Huang HC, et al (2026). Kras G12C- and G12D-driven lung cancers differ in oncogenic potency, immunogenicity, and relapse after Kras inhibition in mouse models. Science Translational Medicine 18(835):eadq6647.
Wei et al. (2023) Dissecting embryonic and extraembryonic lineage crosstalk with stem cell co-culture. Cell 186(26):5859-5875.e24.
Zhu M, et al. (2023) Loss of p53 and mutational heterogeneity drives immune resistance in an autochthonous mouse lung cancer model with high tumor mutational burden. Cancer Cell 41(10):1731-1748.e8.
Develop epigenomics-focused deep learning and bioinformatics algorithms.
We develop deep learning and bioinformatics algorithms to analyze DNA methylation, chromatin accessibility, and histone modifications, revealing how epigenetic regulation shapes gene expression and cellular identity. By integrating these features with genetic and transcriptomic data, we aim to identify functional regulatory elements, interpret noncoding variation, and uncover epigenetic mechanisms underlying human disease.
Publication
IGVF Consortium. (2024) Deciphering the impact of genomic variation on function. Nature Sep; 633(8028):47-57.
Lu Z, Xiao X, Zheng Q, Wang W#, Xu L# (2024) Assessing next-generation sequencing-based computational methods for predicting transcriptional regulators with query gene sets. Briefings in Bioinformatics 25(5):bbae366.
Xu L#, Pierce JL, Sanchez A, Chen KS, Shukla AA, Fustino NJ, Stuart SH, Bagrodia A, Xiao X, Guo L, Krailo MD, Shaikh F, Billmire DF, Pashankar F, Bestrashniy J, Oosterhuis JW, Gillis AJM, Xie Y, Teot L, Mora J, Poynter JN, Rakheja D, Looijenga LHJ, Draper BW, Frazier AL, Amatruda JF#. (2023) Integrated genomic analysis reveals aberrations in WNT signaling in germ cell tumors of childhood and adolescence. Nature Communications 4(1):2636.
Lu Z, Xu L#, Wang X#. (2025) BIT: Bayesian Identification of Transcriptional Regulators from Epigenomics-Based Query Region Sets. Nature Communications 16(1):4966.