Chengyu Du

Chengyu Du

he/him

Doctoral Student in Biostatistics, UMass Amherst

Chengyu Du, Biostatistics PhD Student

I am a doctoral student in Biostatistics at the University of Massachusetts Amherst, advised by Dr. Zhengqing Ouyang in the Ouyang Lab. My research sits at the intersection of statistics, machine learning, and computational biology. I develop statistical and computational methods for structured, high-dimensional biological data, including spatial transcriptomics, 3D genome organization from Hi-C data, and RNA structure ensembles.

Methodologically, I am interested in kernel methods, representation learning, high-dimensional inference, scan statistics, and optimization-based approaches for recovering biological structure from sparse and noisy measurements. Previously, I completed an M.S. in Statistics at the University of Illinois Urbana-Champaign, where I worked with Prof. Ruoqing Zhu on statistical learning for high-dimensional biology, and a B.S. in Mathematics and Applied Mathematics at Beijing Normal University.

Education

  • Doctoral Student in Biostatistics, University of Massachusetts Amherst (Aug. 2024 - Present)
  • M.S. in Statistics, University of Illinois Urbana-Champaign (Aug. 2022 - May 2024)
  • B.S. in Mathematics and Applied Mathematics, Beijing Normal University (Sep. 2017 - Jun. 2021)

Research Interests

  • Spatial transcriptomics and spatially smoothed dimension reduction
  • 3D genome organization, Hi-C contact maps, and structural variant detection
  • RNA cotranscriptional folding and structure ensemble inference
  • Kernel methods, RKHS regularization, and representation learning
  • High-dimensional inference, scan statistics, and constrained optimization

Publications / Preprints

  • Chengyu Du, Mengfan Xu. “Conformal-Style Quantile Analyses for Stochastic Bandits.” arXiv preprint arXiv:2605.07115, 2026. [arXiv]
  • Chengyu Du*, Zirui Li*, Boyi Guo, Ruoqing Zhu. “Smoothed Dimension Reduction in Spatial Transcriptomics Through Functional Embedding.” Manuscript in preparation. *Equal contribution.
  • Chengyu Du, Yingtong Peng, Zhengyuan Qu, Chixueyuan Wang. “Measuring Dependence With HSIC.” World Scientific Research Journal, Volume 6, Issue 10, 2020.

Grants

  • PhRMA Foundation Predoctoral Fellowship in Drug Discovery, “AI-Driven 3D Genome Aberration for Epigenetic Drug Targets”. Submitted Apr. 2026; PI; 2-year stipend; under review.

Conference Presentations

  • Poster: “Mapping Cotranscriptional RNA Structure Dynamics Across the Pre-rRNA Transcript.” UMass SPHHS Research Day 2026, Amherst, Massachusetts, May 2026.
  • Poster: “Robust 3D Chromatin Structure Modeling Integrating Activation Function with Model-Based Distance Embedding.” New England Statistical Society Conference 2025, New Haven, Connecticut, Jun. 2025.
  • Poster: “The Effects of Long Working Hours on Obesity Among American Adults.” American Society for Nutrition, NUTRITION 2025 Conference, Orlando, Florida, Jun. 2025.

Research Experience

RNA Cotranscriptional Folding and Structure Ensemble Inference

Research Assistant, UMass Amherst, supervised by Dr. Zhengqing Ouyang (Oct. 2025 - Present)

  • Developed computational pipelines for RNA cotranscriptional folding analysis by integrating Sfold-sampled secondary structures, DMS reactivity profiles, and accessibility estimates.
  • Formulated structure deconvolution as a constrained optimization problem to infer mixture weights over candidate RNA structures while incorporating experimental probing constraints.
  • Designed cross-transcript and local-neighbor regularization strategies to stabilize inferred RNA structure ensembles across transcript positions.

Statistical Detection of Structural Variants from Hi-C Contact Maps

Research Assistant, UMass Amherst, supervised by Dr. Zhengqing Ouyang (Jul. 2025 - Apr. 2026)

  • Developed a scan-statistics framework for detecting structural variants from Hi-C contact maps, combining kernel density estimation, multi-bandwidth rejection regions, and family-wise error rate control.
  • Designed bandwidth-selection and rejection-region refinement strategies to identify translocation and CNV signals while reducing diagonal and coverage-driven artifacts.
  • Implemented scalable R/C++ pipelines for Hi-C matrix construction, KDE-based signal detection, and genomic interval annotation of candidate SV regions.

Robust 3D Chromatin Reconstruction via Improved MDE

Research Assistant, UMass Amherst, supervised by Dr. Zhengqing Ouyang (Nov. 2024 - Jun. 2025)

  • Developed improved multidimensional embedding methods for reconstructing 3D chromatin structure from noisy Hi-C contact maps with robust losses, weighted distance constraints, and coverage-aware contact transformations.
  • Built automated R/Python pipelines for simulation, parameter tuning, and visualization of chromatin reconstruction accuracy.

Spatially Smoothed Dimension Reduction

Research Assistant, UIUC, supervised by Dr. Ruoqing Zhu (May 2023 - May 2025)

  • Developed spatially regularized embeddings for single-subject and multi-subject spatial transcriptomics settings.
  • Used reproducing kernel Hilbert space regularization to control embedding smoothness.
  • Incorporated interchangeable likelihood functions and preprocessing methods to improve robustness across data sources.

Teaching Experience

University of Massachusetts Amherst

Teaching Assistant (2025 - 2026)

  • BIOSTATS 601: Probability and Statistical Inference for Health Data Science (Spring 2026)
  • BIOSTATS 530: Introduction to Statistical Computing in Data Science using R (Spring 2026 and Fall 2026)
  • BIOSTATS 590B: Introduction to Bayesian Inference (Spring 2026)
  • BIOSTATS 540: Introductory Biostatistics (Fall 2025)

University of Illinois Urbana-Champaign

Teaching Assistant (Fall 2023)

  • STAT 432: Basics of Statistical Learning

Selected GitHub Projects

  • CS690U Final Project: evaluated k-mer frequency features and foundation-model embeddings for clustering biological sequence datasets, including E. coli noncoding RNAs and MopB family proteins.
  • BIOSTAT750 Final Project: compared statistical learning methods for credit-card default prediction using R, cross-validation, ROC analysis, regularized regression, tree-based models, and support vector machines.

Honors & Awards

  • SPHHS Dean’s PhD Fellowship, UMass Amherst, 2024 - 2026 ($30,000)
  • Merit-based Scholarship, Beijing Normal University, 2018 - 2021

Service

  • Subreviewer for ISMB/ECCB 2025 and ISMB 2026
  • Member, American Statistical Association (ASA)
  • Member, New England Statistical Society (NESS)

Skills

  • Programming Languages: Python, R, C++, MATLAB
  • Methods and Tools: statistical machine learning, high-dimensional data analysis, optimization, kernel methods, scan statistics, reproducible research pipelines
  • Typesetting and Publishing: LaTeX, Markdown
  • Developer Tools: VS Code, shell scripting, GitHub
  • Languages: Mandarin (native), English (fluent)