Abstract 18349: Molecular Classification of the Heart Transplant Endomyocardial Biopsy
Background: Endomyocardial Biopsy (EMB) is the standard method to diagnose allograft rejection post HTx. While it is used to support medical decisions, insufficient diagnostic accuracy constitutes a fundamental limitation. The aim of this study is to develop a methodology that improves the classification of the EMB through a non-supervised evaluation of intramyocardial gene expression.
Methods: Sixty-four heart tissues from 47 HTx recipients were subjected to genome wide mRNA sequencing. An unsupervised algorithm using optimal transport to mitigate batch effects and to filter confounding sources of variability was developed to identify molecular signatures of rejection. Linear Mixed Model identified genes statistically significant among the histology defined rejection groups. Weighted Gene Correlation Network Analysis (WGCNA) was used to establish 13 eigengene modules and module-clinical phenotype relationships. Gene Ontology was used for interpretation of the modules in their biological context.
Results: Our algorithm best classified the EMBs into 4 unsupervised clusters solely based on their gene expression. Statistical analysis showed a set of genes differentially expressed among groups defined by histology criteria. Top ranked genes were CLNK, TNFRSF10A, TRADD, CD2, and HLA-A. WGCNA revealed best trait-module correlation was observed between the classes defined by the unsupervised algorithm developed in this study followed by Histology. Figure 1 shows Module-Trait relationships, strength of association, significance and enriched biological process.
Conclusion: We have developed an unsupervised algorithm that classifies the EMBs into 4 functionally distinctive categories. These categories are highly correlated with genomic modules defined by WGCNA and with the clinical phenotypes. To our knowledge, this is the first unsupervised classification of the EMBs. Further validation and performance will be provided at the time of presentation.
Author Disclosures: E. Chang: None. G. Fishbein: None. M. Bakir: None. G. Bondar: None. N. Jackson: None. D. Liem: None. S. Litovsky: None. J. Tallaj: None. C. Starling: None. P. Ping: None. E. Reed: None. M. Deng: None. E. Tabak: None. M. Cadeiras: None.
- © 2015 by American Heart Association, Inc.