Data-driven approach to integrating genomic and behavioral preclinical traumatic brain injury research

J. Russell Huie, Jessica L. Nielson, Jorden Wolfsbane, Clark R. Andersen, Heidi M. Spratt, Douglas S. DeWitt, Adam R. Ferguson, Bridget E. Hawkins

Research output: Contribution to journalArticlepeer-review

Abstract

Understanding recovery from TBI is complex, involving multiple systems and modalities. The current study applied modern data science tools to manage this complexity and harmonize large-scale data to understand relationships between gene expression and behavioral outcomes in a preclinical model of chronic TBI (cTBI). Data collected by the Moody Project for Translational TBI Research included rats with no injury (naïve animals with similar amounts of anesthetic exposure to TBI and sham-injured animals), sham injury, or lateral fluid percussion TBI, followed by recovery periods up to 12 months. Behavioral measures included locomotor coordination (beam balance neuroscore) and memory and cognition assessments (Morris water maze: MWM) at multiple timepoints. Gene arrays were performed using hippocampal and cortical samples to probe 45,610 genes. To reduce the high dimensionality of molecular and behavioral domains and uncover gene–behavior associations, we performed non-linear principal components analyses (NL-PCA), which de-noised the data. Genomic NL-PCA unveiled three interpretable eigengene components (PC2, PC3, and PC4). Ingenuity pathway analysis (IPA) identified the PCs as an integrated stress response (PC2; EIF2-mTOR, corticotropin signaling, etc.), inflammatory factor translation (PC3; PI3K-p70S6K signaling), and neurite growth inhibition (PC4; Rho pathways). Behavioral PCA revealed three principal components reflecting the contribution of MWM overall speed and distance, neuroscore/beam walk, and MWM platform measures. Integrating the genomic and behavioral domains, we then performed a ‘meta-PCA’ on individual PC scores for each rat from genomic and behavioral PCAs. This meta-PCA uncovered three unique multimodal PCs, characterized by robust associations between inflammatory/stress response and neuroscore/beam walk performance (meta-PC1), stress response and MWM performance (meta-PC2), and stress response and neuroscore/beam walk performance (meta-PC3). Multivariate analysis of variance (MANOVA) on genomic–behavioral meta-PC scores tested separately on cortex and hippocampal samples revealed the main effects of TBI and recovery time. These findings are a proof of concept for the integration of disparate data domains for translational knowledge discovery, harnessing the full syndromic space of TBI.

Original languageEnglish (US)
Article number887898
JournalFrontiers in Bioengineering and Biotechnology
Volume10
DOIs
StatePublished - Jan 10 2023

Bibliographical note

Funding Information:
This study was supported in part by the Darrell K Royal Research Fund for Alzheimer’s Disease (BH), Mission Connect, a Program of TIRR (BH), Moody Project for Translational TBI Research (DD and DP), and NIH BD2K RoAD-Trip Fellowship (BH). This work was also supported by the NIH/NINDS under R01NS122888, UH3NS106899, and U24NS122732 to AF and Department of Veterans Affairs under 1I01RX002245 and I01RX002787 to AF.

Publisher Copyright:
Copyright © 2023 Huie, Nielson, Wolfsbane, Andersen, Spratt, DeWitt, Ferguson and Hawkins.

Keywords

  • TBI
  • behavior analysis
  • data-driven learning
  • genomic
  • multivariate statistical analyses
  • traumatic brain injury

PubMed: MeSH publication types

  • Journal Article

Fingerprint

Dive into the research topics of 'Data-driven approach to integrating genomic and behavioral preclinical traumatic brain injury research'. Together they form a unique fingerprint.

Cite this