Identification of alternative splicing regulatory patterns and characteristic splicing factors in heart failure using RNA-seq data and machine learning

Abstract

Heart failure (HF) represents the advanced stage of several cardiovascular disorders. This study aimed to build an alternative splicing regulatory network and identify potential splicing factors involved in HF utilizing RNA-seq data and machine learning algorithms. We performed bioinformatics analysis on RNA-seq datasets containing samples from HF patients and normal individuals to obtain gene expression matrices and identify differently regulated alternative splicing events in HF. By calculating percent spliced-in (PSI) value, we identified 4055 abnormal alternative splicing events of 3142 genes in HF. These genes were significantly enriched in PPAR signaling, regulation of actin cytoskeleton, and muscle contraction. Interestingly, based on abnormal alternative splicing events, two distinct clusters of HF patients with distinct molecular mechanisms and pathways were identified using unsupervised clustering. Additionally, we built a regulatory network consisting of heart failure-related alternative splicing and splicing factors. Subsequently, we identify 203 HF specific pairs between splicing factors and alternative splicing events. Four splicing factors (RBM5, ZRANB2, HnRNPF, and HnRNPA0) were found using LASSO and SVM-RFE algorithms, their expression patterns were confirmed in two other microarray datasets. Our study clarifies involvement of splicing factors and alternative splicing events in HF by thoroughly analyzing RNA-seq data with machine learning methods. The findings may advance our understanding of the regulatory systems underlying biological processes associated with heart failure by providing candidates for further investigation and markers for diagnostic and therapeutic purposes.

Keywords

• Alternative splicing
• Heart failure
• Characteristic splicing factor
• Machine learning