Academic year 2023-2024 seminars

Academic Year 2023-24 Seminars

September 15, 2023

Speaker: Joseph Ibrahim, University of North Carolina at Chapel Hill

Title: The Scale Transformed Power Prior for Time-to-Event Data

Poster: Ibrahim_Poster 20230915

Abstract: In clinical trials, data is often available from a previous trial with a different outcome (i.e., binary vs time-to-event). The power prior proposed by Ibrahim and Chen (2000) does not account for different data types in the context discussed here. To accommodate settings in which the historical data and the current data involve different data types, we develop the partial-borrowing scale transformed power prior (straPP) for several commonly used time-to-event models. The partial-borrowing straPP is developed through rescaling the parameter vector from the historical data to align with that of the new data using a transformation based on the Fisher information matrices from the two data models. We also develop the generalized scale transformed power prior (Gen-straPP) to provide added robustness for the case in which the scaled parameters are not equal. Several real data sets from the Eastern Cooperative Oncology Group are used to motivate the use of the proposed methods. We demonstrate the advantages of the partial-borrowing straPP over other common priors via simulation and real data analyses using the proportional hazards model and the mixture cure rate model.

September 22, 2023

Speaker: Casey Taylor, Johns Hopkins University

Title: Clinical decision support for unsolicited genomic results.

Poster: Taylor Poster 20230922

Abstract: Given that clinical genomic tests can be initiated outside of the clinical setting (for example, in a research study), from the clinician’s perspective, they can be characterized as “unsolicited,” which brings the challenge of how to determine the value and use of those data in patient care. Dr. Taylor will describe her ongoing research investigating the information and technical requirements for software to enable risk-benefit stratification of clinical decision support (CDS) for unsolicited genomic results (UGR) and the attributes needed for genomic service providers to decide how to prioritize and implement CDS based on UGR. Furthermore, Dr. Taylor is exploring how approaches used for UGR apply for prediction models. The broader impacts of this work are enhancing the productivity and effectiveness of genomic service providers; a greater awareness of how software can be used to support the work of genomic service providers; and an increased ability to implement CDS based upon UGR and prediction models within heterogeneous clinical IT infrastructures.

September 29, 2023

  • Speaker: Hongtu Zhu, University of North Carolina at Chapel Hill
  • Title: Establishing the Causal Genetic Imaging Clinical Pathway for Brain-Related Disorders
  • Abstract: Our causal genetic imaging clinical (CGIC) pathway is motivated by the joint analysis of genetic, imaging, and clinical (GIC) data collected in many large-scale biomedical studies, such as the UK Biobank study and the Alzheimer’s Disease Neuroimaging Initiative (ADNI) study. We propose a statistical framework based on partially functional linear regression models and varying coefficient models to map CGIC pathway for phenotypes associated with brain-related disorders. We develop a joint model selection, estimation, and inference procedure by embedding imaging data in the reproducing kernel Hilbert space and imposing the penalty for the coefficients of scalar variables. We systematically investigate the theoretical properties of scalar and functional efficient estimators, including non-asymptotic error bound, minimax error bound, and asymptotic normality. We apply the proposed method to the ADNI and UKB datasets to identify important features from several millions of genetic polymorphisms and study the effects of a certain set of informative genetic variants and the hippocampus surface on different clinical variables.

October 6, 2023

Speaker: Claudia Solis-Lemus, Department of Plant Pathology, Affiliate Faculty Biostatistics and Medical Informatics

Title: Inferring networks

Poster: 20231006Claudia Solis-Lemus_Poster

Abstract:Networks are graphical structures that appear in a variety of biological applications from phylogenetic networks to study evolution to interaction networks to study microbial communities in soil and plants. I will describe the novel statistical advances (and challenges) to estimate 1) phylogenetic networks from genome-wide data, and 2) microbial networks from abundance data. I will conclude with some examples of how deep learning models can be applicable to infer these types of networks.

October 13, 2023

Speaker: Solomon Harrar, University of Kentucky

Title: Nonparametric finite mixture: applications in overcoming misclassification bias

Poster: 20231013Harrar_Poster

Abstract: In precision medicine, patients are classified into disease subtypes using biomarkers. The evaluation of treatments for the subtypes traditionally assumes that the biomarkers are accurate, which is seldom the case. The impact of diagnostic misclassification in statistical inference has been recently investigated in parametric model contexts and shown to introduce severe bias in estimating treatment effects and give grossly inaccurate inferences. The research presented in this talk aims to address these problems in a fully nonparametric setting. Methods for consistently estimating and testing meaningful yet nonparametric treatment effects are developed. Along the way, we construct estimators for misclassification error rates and investigate their asymptotic properties. The proposed methods are applicable for outcomes measured in ordinal, discrete, or continuous scales. They do not require any assumptions, such as the existence of moments.  Simulation results show significant advantages of the proposed methods in bias reduction, coverage probability, and power. The applications of the proposed methods are illustrated with gene expression profiling of bronchial airway brushing in asthmatic and healthy control subjects.

October 20

Speaker: Yeonhee Park, Department of Biostatistics and Medical Informatics

Title: Data-driven monitoring for phase II clinical trial designs based on percentile event time test


Poster: 20231020Park_Poster

Abstract: The goal of phase II clinical trials is to evaluate the therapeutic efficacy of a new drug. Some investigators want to use the time-to-event endpoint as the primary endpoint of the phase II study to see the improvement of the therapeutic efficacy of a new drug in median survival time. Recently, median event time test (METT) has been proposed to provide a simple and straightforward rule which compares the observed median survival time with the prespecified threshold. However, median survival time would not be observed during the trial if the drug performs well and indeed cures most patients or if the accrual rate is so fast. To address the issues in clinical practice, we first propose a percentile event time test (PETT), which generalizes METT to any percentile of the survival time, and develop data-driven monitoring for phase II clinical trial designs based on PETT. We illustrate the proposed method with a trial example and evaluate the performance of the method through simulations.


Special Event – SMPH Collaborate

Location: 1306 Health Sciences Learning Center

More Information and Registration:

October 26 – DeMets Lecture

(NOTE VENUE and TIME CHANGE: HSLC 1335 from 3:30-4:45 pm, followed by a reception in the HSLC Atrium from 4:45-6:00 pm.)

Speaker: Colin Begg, Memorial Sloan Kettering Cancer Center


October 27

Speaker: Alan Moses, University of Toronto


November 3

Speaker: Lu Mao, Biostatistics and Medical Informatics

NOTE: Alternate Location: 1345 HSLC


November 10

Speaker: Dongjun Chung, Ohio State University


November 17

Speaker: Christina Leslie, Memorial Sloan-Kettering Cancer Center


November 24

  • no seminar (holiday)

December 1

  • Speaker: Felix Elwert, Department of Sociology, Faculty Affiliate Department of Biostatistics and Medical Informatics
  • NOTE: Alternate Location: 1306 HSLC
  • Zoom:

Thursday December 7, 12-1pm

  • Speakers: Fall 2023 Student Rotation Presentations
  • Biotech Auditorium
  • Zoom:

Friday December 8, 12-1pm

  • Speakers: Fall 2023 Student Rotation Presentations
  • Biotech Auditorium
  • Zoom:

Friday, January 26, 2024

  • Speaker: Keaven M. Anderson, PhD, Scientific AVP, Methodology Research, Merck & Co., Inc.
  • Biotech Center Auditorium
  • Zoom Link: Https://
  • Title: Weighted Parametric Multiple Test Procedures in Group Sequential Design (WPGSD)
  • Abstract: There are often multiple questions that a Phase 3 clinical trial will attempt to answer to justify marketing· approval for a new treatment. Questions that may need to be answered include evaluation of multiple experimental treatment groups, what population is most appropriate for treatment, and/or what endpoints are impacted. In addition, some, questions can be answered early in a trial (e.g., reduction in tumor size) and others may take much longer (e.g., improvement in overall survival). Interim analyses can be useful both for futility and for stronger than expected efficacy benefit. In this contest, we consider strong control of Type I error in group sequential trials testing multiple hypotheses. We extend graphical methods for multiplicity control in group sequential trials to account for correlations between hypothesis tests. We also consider fully-adjusting p-values for both multiple hypothesis testing and group sequential testing; this enables simple comparison of p-values to an overall family-wise Type I error when evaluating statistical significance. Practical issues are considered along with gains in efficiency that might be expected.

February 9, 2024

Speaker: Carolyn Uhler, PhD, Massachusetts Institute of Technology

Title: Multimodal Data Integration: From Biomarkers to Mechanisms



Poster: 0209-Uhler-poster-2

Abstract: An exciting opportunity at the intersection of the biomedical sciences and machine learning stems from the growing availability of large-scale multi-modal data (imaging-based and sequencing-based, observational and perturbational, at the single-cell level, tissue-level, and organism-level). Traditional representation learning methods, although often highly successful in predictive tasks, do not generally elucidate underlying causal mechanisms. I will present initial ideas towards building a statistical and computational framework for causal representation learning and its applications towards identifying novel disease biomarkers as well as inferring gene regulation in different disease contexts.

Friday, February 16, 2024

Speaker: Haohan Wang, Assistant Professor, School of Information Science, UIUC

Title: Advancing Precision Medicine: Tailored Genomic Insights and AI-Driven Automation in Complex Disease Research.


Poster: 0216-Wang-poster.docx

Abstract: Understanding the genetic basis of complex human disorders is crucial in the realm of medical research, particularly in the context of precision medicine. Complex diseases such as Alzheimer’s and various forms of cancer exhibit significant genetic heterogeneity, posing challenges in diagnosis and treatment. This talk will focus on two cutting-edge research projects that integrate genomics and machine learning, driving forward the field of precision medicine. The first part of the presentation introduces the Heteroscedastic Personalized Regression (Het-PR) model, a new approach in genomic analysis for Alzheimer’s disease. Het-PR diverges from conventional models by emphasizing personalized treatment trajectories based on individual genetic factors. This model is adept at pinpointing unique genetic mutations and influential genetic factors in patients, which traditional models often miss due to their generalized approach. Demonstrated to be more effective in both simulation and real-world scenarios, Het-PR enhances the capability to devise personalized treatment plans for Alzheimer’s patients, taking into account their individual genetic backgrounds. In the second part, the focus shifts to the Team of AI-made Scientists (TAIS), an innovative initiative using Large Language Models (LLMs) to automate the scientific discovery process in disease research. TAIS signifies a paradigm shift in genomic analysis, moving from traditional, labor-intensive methods to a more automated, efficient, and scalable model. This AI-driven team simulates various roles typically found in research teams, working collaboratively to autonomously perform complex tasks like data analysis and gene identification. The application of TAIS in identifying genes predictive of disease status exemplifies its potential to markedly accelerate the pace of discovery in medical genomics.


February 23, 2024

Speaker: Tim Millers, Boston Children’s Hospital and Harvard Medical School

Title: Bringing Biomedical NLP into the Large Language Model Era

Zoom:  https://uwmadison.zoom,us/j/95515534304

Poster: 0223-Miller-poster.docx

Abstract: Biomedical NLP aims to make it easier to extract important information from unstructured texts like electronic health records, biomedical journal articles, regulatory documents, etc, and to use this information to improve our lives. In this talk, we will describe recent work in this area from Dr. Miller’s lab, and connect it to bigger questions facing the field of biomedical NLP with the emergence of powerful class of models known as large language models (LLMs). How important is dataset creation? Will NLP experts and subject matter experts need each other anymore? Will LLMs still suffer from out-of-domain performance loss as supervised models?

March 1 – Chaowei Ziao, UW-Madison – Cancelled

March 8 – Dave Zhao, University of Illinois Urbana-Champaign

March 15

Speaker: Jingwen Yan, Indiana University, Indianapolis

Title: Integrative -0mics for improved understanding and risk stratification of Alzheimer’s disease.

Zoom: https://uwmadison.zoom.uw.j.95515534304

Poster: 0315-Yan-poster.docx

Abstract: Integrative -omics is an emerging research field that aims to extract the knowledge from the broad multi-omic data landscape.  While multiple domains included, such as brain imaging, genetics, transcriptomics and proteomics, it offers great promise to illuminate the causal pathway from genotype to phenotype and to provide optimal molecular phenotypes for early therapeutic intervention. My research has been focused on development and application of computational approaches 1) to explore the complementary information between –omics data and 2) to investigate how that could benefit the biomarker discovery and early detection of Alzheimer’s disease (AD). In this talk, I will introduce some of our recent work on exploration of AD molecular mechanism and risk stratification with integrative -omics approaches.

March 22 – Sanjib Basu, University of Illinois-Chicago

March 29 – no seminar

April 5 – Majid Afshar, UW-Madison

April 12 – Jiang Bian, University of Florida Health

April 19 – Bo Wang, University of Toronto

April 25 – Agni Orfanoudaki, Massachusetts Institute of Technology

April 26 – Yuan Ying, University of Texas MD Anderson Cancer Center

May 3, 2024

  • Title: Biomedical Data Science Program Spring 2024 Student Rotation Presentations
  • Time: 11:45-1:00 pm
  • Zoom:
    Meeting ID: 912 4566 6497

  • Poster: Spring 2024 Rotation Presentations
  • Presentations:
    • Aurod Ounsinegad
      • Ranking and Clustering Counties in Washington State Based on Health Outcomes
      • Rotation Mentor: Professor Ron Gangnon
    • John Peters
      • Channel Adaptive Transformers for Biomedical Imaging
      • Rotation Mentor: Professor Juan Caicedo
    • Jifan Gao
      • Automated stratification of trauma injury severity across multiple body regions using multi-modal, multi-class machine learning models.
      • Rotation Mentor: Professor Majif Afshar
    • Tim Gruenloh
      • Benchmarking Multimodal Machine Learning Methods in Omics Data
      • Rotation Mentor: Professor Daifeng Wang
    • Xhongxuan Sun
      • Network QTL Analysis with Population-Scale scRAN-seq Data
      • Rotation Mentor: Professor Sunduz Keles
    • Elliot Xie
      • Finding Causal Relationship Between Clonetype Size and Mutation Burden Using Paried Single-Cell and TCR Data
      • Rotation Mentor: Professor Michael Newton
    • Lingxin Cheng
      • Title tbd
      • Rotation Mentor: Professor Christina Kendziorski

Completed seminars:

September 8, 2023

Title: Summer 2023 Student Rotation Presentations

Ryan Kassab
Project: Statistical Methods for Analyzing Stepped Wedge Cluster Randomized Trials: A Selective Review
Mentor: SushmitaRoy

Yujia Cai 
Project: Detecting Intron Retention Events in DEAD-Box Helicase 41 Germline Variants
Mentor: SunduzKeles

Tinghui Xu
Project: Using Optimal Transport for Quantile Treatment Effect Estimation
Mentor: MenggangYu

Sierra Strutz
Project: Predicting Ovarian Cancer Using EHR Data
Mentor: Irene Ong

Emma Croxford
Project: Determining Potential Differences in Secretome Cytokine Expression using Linear Discriminant Analysis”
Mentor: RickChappell

Jie Sheng
Project: Evaluating Cell-Cell Interaction (CCI) Scores Using Spatial Transcriptomics Data
Mentor: Huy Dinh

Yuda Liu
Project: Benchmarking Causal Graph Inference Algorithms on Simulated and Real Expression Datasets
Mentor: SushmitaRoy