My research on multiplexed tissue imaging data aims to exploring the spatial cell arrangement of tumor microenvironments in melanoma and bladder cancer patients. I developed a comprehensive analysis pipeline for the whole-slide multiplexed immunofluorescence imaging data and proposed SpaTopic, a highly scalable spatial topic model that integrates cell phenotypes with their spatial arrangement to recover the spatial dynamics of cells and infer spatial tissue architecture. Through analyzing pre-treatment whole-slide tissue sections from 53 patients with unresectable melanoma treated with anti-PD-1, our work identified B cell aggregates near tumors as potential biomarkers for cancer immunotherapy.
Peng X, Smithy J, Yosofvand M, Kostrzewa C, Bleile M, Ehrich F, Lee J, Postow M, Callahan M, Panageas K, Shen R, “Scalable topic modelling decodes spatial tissue architecture for large-scale multiplexed imaging analysis”. Nature Communications 16, 6619 (2025). [Link]
Smithy J*, Peng X*, Ehrich F, Moy A, Aleynick N, Li Y, Maher C, Lee J, Panageas K, Hollman T, Callahan M, Shen R. Quantitatively defined stromal B cell aggregates are associated with response to checkpoint inhibitors in unresectable melanoma. Cell Reports, 2025. 44(4), 115554. [Link]
Yosofvand M, Edmiston S, Smithy J, Peng X, Kostrzewa C, Lin B, Ehrich F, Reiner A, Miedema J, Moy A, Orlow I, Postow M, Panageas K, Seshan V, Callahan M, Thomas N, Shen R. “Spatial Immunophenotyping from Whole-Slide Multiplexed Tissue Imaging Using Convolutional Neural Networks”. (Submitted) [Link]
My previous postdoctoral research at Memorial Sloan Kettering Cancer Center mainly focused on identifying biomarkers from longitudinal flow cytometry to predict outcomes for cancer patients undergoing immunotherapy. I developed TopicFlow, a computational framework that adapts topic modeling from natural language processing to analyze longitudinal single-cell data. Applying this framework to a large cohort of melanoma patients, I uncovered dynamic T-cell compositions related to treatment resistance and toxicity. Our collaboration work has identified a regulatory T cell subset potentially crucial for response to anti-PD-1 and anti-LAG-3.
Peng X, Lee J, Adamow M, Maher C, Postow M, Callahan M, Panageas K, Shen R. A topic modeling approach reveals the dynamic T cell composition of peripheral blood during cancer immunotherapy. Cell Reports Methods. 2023; 3(8):100546. PMC10475788 [Link]
Rolig A*, Peng X*, Sturgill E, Mick C, Mcgee G, Kasiewicz M, Miller W, Koguchi Y, Kaufmann J, Yanamandra N, Griffin S, Smothers J, Garnet-Benson C, Jarchem I, Adamow M, Lee J, Shen R, Callahan M, Redmond W. The response to anti–PD-1 and anti–LAG-3 checkpoint blockade is associated with regulatory T cell reprogramming.Science Translational Medicine. 17, eadk3702(2025). [Link]
Peng X, Ehrich F, Shen R, Katherine P. Unlocking Biomarkers for Cancer Immunotherapy: Immune Monitoring with High-Parameter Flow Cytometry. Journal of Cancer Immunology. 2024; 6(2):51-54. [Link]
During my Ph.D. at Iowa State University, I developed advanced bioinformatics algorithms focused on denoising and bias correction for next-generation sequencing data, particularly in distinguishing true biological variants from errors caused by PCR amplification and sequencing—a critical issue in microbiome research. I created a reference-free, model-based clustering algorithm, AmpliCI, that outperforms existing methods in resolving single-nucleotide variants in large Illumina amplicon datasets. Also it is computationally efficient with 7 min for clustering 2.0M sequences. Additionally, I addressed abundance distortions from PCR bias by developing a probabilistic framework using Unique Molecular Identifiers (UMIs) to accurately deduplicate sequences and resolve true biological amplicon sequences.
Peng X, Dorman KS. AmpliCI: a high-resolution model-based approach for denoising Illumina amplicon data. Bioinformatics. 2021; 36(21): 5151-5158. PMC7850112. [Link]
Peng X, Dorman KS. Accurate estimation of molecular counts from amplicon sequence data with unique molecular identifiers. Bioinformatics. 2023;39(1). btad002. PMC9891248. [Link]
Dorman K, Peng X, Zhang Y. Denoising Methods for Inferring Microbiome Community Content and Abundance. Book chapter for “Statistical Analysis of Microbiome Data” in the Frontiers in Probability and the Statistical Sciences series by Springer. 2021;:13-25. [Link]