Background: N1-methyladenosine (m1A) is a reversible post-transcriptional modification in mRNA, which has been proved to play critical roles in various biological processes through interaction with different m1A regulators. There are several m1A regulators existing in the human genome, including YTHDF1-3 and YTHDC1.
Methods: Several techniques have been developed to identify the substrates of m1A regulators, but their binding specificity and biological functions are not yet fully understood due to the limitations of wet-lab approaches. Here, we submitted the framework m1ARegpred (m1A regulators substrate prediction), which is based on machine learning and the combination of sequence-derived and genome-derived features.
Results: Our framework achieved area under the receiver operating characteristic (AUROC) scores of 0.92 in the full transcript model and 0.857 in the mature mRNA model, showing an improvement compared to the existing sequence-derived methods. In addition, motif search and gene ontology enrichment analysis were performed to explore the biological functions of each m1A regulator.
Conclusions: Our work may facilitate the discovery of m1A regulators substrates of interest, and thereby provide new opportunities to understand their roles in human bodies.
Keywords: m1A; machine learning; substrate.
© 2022 The Author(s). Published by IMR Press.