Identification of novel RNA-sequencing biomarkers for early detection of heart transplantation rejection utilizing machine learning techniques

Abstract Introduction Heart transplantation (HTx) remains the primary therapeutic intervention for patients with end-stage heart failure; however, it is associated with various complications, notably the rejection of allograft tissue that occurs in about 30% of patients during the first year following HTx and subsequent mortality. The gold standard method for diagnosis of rejection is endomyocardial biopsy (EMB), which requires hospitalization and is an invasive method. Therefore, the exploration of non-invasive alternatives to EMB including molecular biomarkers is of paramount importance. Purpose In this work, we integrated bioinformatics and machine learning (ML), a subfield of artificial intelligence (AI), to investigate new and non-invasive biomarkers for the early detection of HTx rejection. Material and Methods Here, the gene expression omnibus (GEO) was used to download microarray data for analysis and validation. Subsequently, R programming was applied to determine the differentially expressed genes (DEGs) of samples of rejection vs no rejection. Finally, the biomarkers were separated into two training and testing groups and identified using deep learning, an ML technique. Results Following downloading data from 137 samples, the adjusted t-test and p-values were used to compare the candidate genes. Based on the feature importance score, the top ten genes (TYMS, WARS, AIM2, CXCL9, TRAT1, HLA-DRB3, TNFRSF9, GZMH, IL-32, and AIF1) were chosen as candidates for additional analysis, and expression of each gene had a notable rise in the rejection group. The top three genes for HTx rejection detection among these ten were CXCL9, TNFRSF9, and GZMH. The novel biomarkers for diagnosis are GZMH and TNFRSF9. However, CXCL9 was previously investigated as diagnostic gene for HTx rejection. Conclusion TNFRSF9 is a member of the tumor necrosis factor (TNF) receptor superfamily and is expressed by various cell types, including endotheliocytes, natural killer (NK) cells, and activated CD8+ and CD4+ T lymphocytes. This receptor plays a critical role in regulating apoptosis, cellular proliferation, and cell survival. The activity of CD8+ T cells is significantly regulated by TNFRSF9. Granzyme H (GZMH) is a protease that is also involved in phagosome formation, cytokine regulation, and the remodeling of the extracellular matrix. Upon the detection of target cells, perforins are secreted and integrate into the target cell membrane, enabling granzymes to enter and induce apoptosis. While Granzyme A has been previously noted for its strong expression during HTx rejection, our current findings indicate that GZMH is also highly expressed in this situation. CXCL9, which belongs to the CXC subfamily of chemokines, has emerged as a non-invasive biomarker for HTx rejection, corroborating the outcomes of our investigation.