The Athey Lab

4. Enhancing Academic Productivity through Generative AI-Powered Workflows (Lars Fritze and Brian Athey) 

In today’s academic landscape, enhancing research output and streamlining the writing process are essential. In this project, we aim to leverage Generative Artificial Intelligence (GenAI) technology to improve the creation of academic manuscripts and grant proposals. By developing and refining specialized Generative Pre-trained Transformers (GPTs), we strive to support researchers at various stages of their writing projects, reducing the time spent on academic writing and facilitating more effective research communication. 

Utilizing both accessible platforms and proprietary tools, we plan to establish workflows that draw upon extensive curated knowledge bases and the latest large language models (LLMs). Our goal is to guide colleagues and aspiring PIs through the intricate research documentation process, from ideation to submission. These knowledge bases, tailored to specific projects, domains, and tasks, ensure the workflows are relevant and practical. 

In addition, we are committed to showcasing detailed examples and step-by-step guides. These resources aim to illustrate how GenAI-powered workflows can transform literature review and the initial stages of drafting research manuscripts or grants, addressing common challenges such as writer's block.  

Ultimately, while the responsibility for research integrity remains with the individual researcher, the judicious use of GenAI tools can significantly enhance the efficiency and quality of academic writing. In doing so, we aim to foster a more dynamic and collaborative research environment at the University of Michigan, propelling our community toward discoveries and innovations. 

5. Integrating Genomics and Health Data to Refine Hypertension Management: Insights from the NIH All of Us Research Program cohort (Lars Fritze and Brian Athey) 

Nearly half of the U.S. adult population (48.1%, or approximately 119.9 million) have hypertension, and only about one in four of these individuals manage to control their condition. Uncontrolled hypertension is associated with a high risk of cardiovascular diseases and was a primary or contributing cause of over 691,000 deaths in the U.S. in 2021 alone (CDC 2021). 

This project aims to identify how comorbidities, drug-drug interactions, and genetic variations affect hypertension medication effectiveness within the All of Us cohort. By utilizing genomic data, harmonized electronic health records, and survey data on social determinants of health, we aim to identify patterns and disparities in blood pressure control across this diverse participant group. 

We plan to use mixed models to estimate individual blood pressure baselines and trajectories over time. Additionally, we aim to integrate genetic predictors of drug response by utilizing known pharmacogenetic variants and tools like Aldy or Stargazer along with All of Us's genomics data. This integration will improve our comprehension of pharmacogenetic phenotypes affecting drug efficacy and safety. 

The anticipated outcomes include insights into the interplay between genetics, comorbidities, and drug responses, facilitating the identification of patient subgroups with distinct treatment profiles. This research aims to study the feasibility of analyzing retrospective biobank data to understand blood pressure control variations. Through this approach, we strive to demonstrate the potential of biobank data in informing future research directions and methodologies in hypertension management for more targeted and effective interventions. 

6. NIH VIOLIN 2.0: Vaccine Information and Ontology Linked Knowledge-base 

Major Goals: To promote community-wide data/metadata standardization and analysis, advance the understanding of the vaccine mechanisms, and support rational vaccine mechanism study and rational vaccine design against various infectious diseases, leading to safer public health. 

Project Number: NIH-NIAID U24A171008
Name of PD/PI: He, Oliver, Co-I, Athey 

7. NIH University of Michigan O’Brien Kidney Translational Core Center 

Major Goals: Supports the translational pipeline providing resources to investigators in our Institutional and International Research Bases: 1.) Expands our unique CKD cohort combined with its longitudinal tissue, urine and serum biobanks to allow our research base investigators to investigate the molecular causes and endpoints of chronic kidney diseases; 2.) Disseminates and supports modern and powerful systems biological approaches for investigators to help them identify novel and robust biomarkers, endpoints and targets for diagnosis and treatment of CKD; 3.) Expert analysis and integration of cohort and systems data for our investigators using sophisticated bioinformatics and database integration that promote identification of specific pathways and targets for treatments for individuals or groups of subjects with CKD. 

Note: I am not involved with individuals, funding, or data from outside countries on this grant. 

Trainees and faculty participating in this project may be from various countries. None is currently paid under my supervision. 

Status of Support: Active
Project Number: P30 DK08194325-01A1
Name of PD/PI: Pennathur, Sub; Co-I Athey 

8. High-Performance Computing Cluster for Biomedical Research 

NIH NIGMS 1S10OD0268 

Name of PD/PI: PI Athey 

We have recomposed the S10 advisory committee to include the following members, intended to provide a balance of stakeholders in terms of research interests, usage patterns, and organizational homes: 

Prof. Lydia Freddolino (chair) -- Dr. Freddolino's research group is a major user of the resource, including much of the method development and application for protein structure/function prediction.  

Prof. Brian Athey -- Dr. Athey is chair of the Department of Computational Medicine and Bioinformatics, the administrative home of the S10, and has lab members who are direct users of the resource.  

Prof. Josh Welch -- Dr. Welch is another current major user of the GPU partition of the S10 resource for running deep learning based protein structure prediction.  

Prof. Jianzhi Zhang -- Dr. Zhang's lab is a heavy user of the CPU partition of the S10 resource for analyzing.  

Dr. Paul Wolberg (representing Prof. Denise Kirschner) -- Dr. Wolberg is a research scientist in the Kirschner lab, which makes heavy use of the S10 CPU partition to simulate the dynamics of tuberculosis infection.  

Jonathan Poisson -- Mr. Poisson is a systems administrator in the Department of Computational Medicine and Bioinformatic and assists in running the S10 resource.  

Brock Palen -- Mr. Palen is a systems administrator at the Academic Research Computing office at the University of Michigan and serves as a liaison to other campus-level IT resources.  

9. NIH The role of epithelial barrier dysfunction in food anaphylaxis 

Major Goals: Food allergy affects nearly 10% of the United States population, and misdiagnosis leads to difficult, anxiety provoking, and growth-limiting food avoidance. Current diagnostic tools are fraught with inaccuracy or require ingesting possible food allergens under medical monitoring, which is cumbersome, risky, and costly. The goal of this project is to understand the role of leaky skin and gut barriers in food allergy to develop more accurate and less onerous tests to diagnose food allergy. 

Status of Support: Active
Project Number: F065420
Name of PD/PI: PI Chase Schuler, Co-I, Athey 

10. Title: COMPASS: A comprehensive mobile precision approach for scalable solutions in mental health treatment (Fritshe, Athey, Tewari) 

Major Goals: This study applies machine learning approaches to genomic, active and passive mobile behavioral tracking and EHR data from a large cohort before and during digital and clinic-based mental health care to develop individualized prediction models that will optimize mental health treatments. 

Status of Support: Pending
Project Number: U01 MH136025
Name of PD/PI: Bohnert-Contact/Sen/Fritsche
Source of Support: NIH
Project/Proposal Start and End Date: 04/01/2024-03/31/2029
Primary Place of Performance: University of Michigan, Ann Arbor
Total Award Amount $18,721,603