Eric Rouchka and David Magnusson, University of Louisville, Louisville, USA

Systems integration for predicting behavioral phenotypes and transcriptional response to activity

Funded in: 2020, 2021, 2022


Back to overview

Problem: Machine learning approaches can be used to analyze behavioral and functional data
Target: Identify biological processes, cell-types, and individual genes that are associated with distinct behavioral-based outcomes
Goal: Identify at a molecular level new target transcripts, proteins and mechanisms

Introduction: A devastating feature of spinal cord injury (SCI) is the disruption of peripheral organ innervation and regulation. Neuropathic pain, cardiovascular disease, and liver damage are common in SCI patients, hinder recovery, and affect quality of life. A non-invasive approach to powerfully attenuate SCI-induced complications in a number of cell/tissue compartments is activity/exercise (A/E).  A/E can enhance recovery and ameliorate SCI-induced complications, and can potentially be applied as part of the activities of daily life.  However, little is known about events occurring at the cellular/molecular level contributing to improved phenotypes resulting from A/E.
Problem Statement: The researchers hypothesize that machine learning approaches can be used to analyze behavioral and functional data in order to identify clusters of differences and similarities for subjects/animals with SCI, in particular those exposed to a regimen of A/E.  These phenotypic responses can then be integrated with “-omics” data to identify biological processes, cell-types, and individual genes that are associated with distinct behavioral-based outcomes, acutely and chronically.
Methods: The first aim of this three-aim project will collect, organize, and integrate behavioral, functional, and transcriptional data both from their own studies as well as publicly available datasets in a manner that is Findable, Accessible, Interoperable, and Reusable (FAIR), as well as developing machine learning approaches to pull out additional information from these data sets. The second aim involves an iterative approach for integrating transcriptomic and behavioral data using Topological Data Analysis.  The final aim will produce additional behavioral and transcriptomic data for SCI models incorporating A/E.
Potential Application: As a result of their integration of transcriptomic, behavioral, and functional data, the researchers hope to identify at a molecular level new target transcripts/proteins and mechanisms that can be further pursued for potential therapeutic directions.