Distinct Patterns of Gene Expression Identified Longitudinally and within AlloMap Score Ranges are Associated with Clinical Outcomes

Abstract

Summary of Objectives Gene expression profiling (GEP, AlloMap) has become the standard of care in heart transplant patient management with over 120,000 performed. The large size of the dataset enables identification of distinct patterns of longitudinal and component gene expression. OAR is a prospective observational heart transplant registry with integrated GEP testing to elucidate factors that contribute to long term outcomes. Samples with GEP scores and individual gene data but not included in OAR have been examined by unsupervised methods to identify patterns of gene expression. These patterns can then be correlated with outcomes collected in OAR to identify specific signals that are indicative of different outcomes. Methods Latent class mixed models were used to estimate profiles of GEP score over time with 62,833 samples from 8,066 patients with at least two GEP score measures. Distinct patterns were identified in a 3-class model; with 69% of the 8,066 patients in a class with a gradual increase in GEP score over time; 23% with a stable GEP score, and 8% with a rapid increase in GEP score. The full dataset of 19,761 patients and 104,827 samples was used to identify patterns of gene expression among the 7 gene terms and the principle components contributing to the GEP Scores. PDCD1, ITGA4, RHOU, and SEMA7A were correlated more closely than the remaining terms. Additionally, the 7 terms were used in an archetypal analysis that identified eight unique archetype patterns of gene expression. For example, two archetypes, #4 and #8 have median GEP scores of 29.4 and 29.9, but different gene patterns. Archetype #4 has high RHOU and PF4/G6B1, but low SEMA7A. Archetype #8 has high IL1R2/FLT3/ITGAM and low PF4/G6B1. Endpoints The principal components analysis identified correlations that match the known and hypothesized biological mechanisms of the genes; the highest correlation was among four genes known or hypothesized to be indicators of T cell activation. The identified archetypal patterns will be assigned to patients in the OAR registry to identify clinical correlates with each archetype group. Among the clinical factors that will be investigated are malignancies, infections, cellular rejection, and cardiac allograft vasculopathy.