Mapping Allele-specific RBP Binding by a Machine Learning Coupled RNA Editing Strategy in Human Embryonic Stem Cells

Abstract RNA-binding proteins (RBPs) play critical roles in regulating the maintenance and differentiation of embryonic stem cells. A critical step in understanding the functions of RBPs is the identification of their bound RNA transcripts. To achieve this, we developed an inducible strategy termed RNA-Editing-Based-RNA-seq (REB-seq) coupled with machine learning to capture the potential RNA targets bound by the RBPs-of-interest. By fusing RBPs to the catalytic domain of ADAR or APOBEC1, which mediates the A-to-I (A-to-G) or C-to-U (C-to-T) editing, respectively, REB-seq enables transcriptome-wide identification of RBP-bound RNAs without high-quality antibodies. Using REB-seq, we characterized the mRNAs bound by the N6-methyladenosine (m6A) readers IGF2BP1, IGF2BP2, and IGF2BP3 in human embryonic stem cells. Coupling REB-seq analysis with machine learning further improves the accuracy of target RNA characterization. Furthermore, REB-seq identifies associated single nucleotide polymorphisms that may affect the RBP binding and imply disease pathogenesis. Collectively, REB-seq coupled with machine learning provides a robust and accessible approach for profiling allele-specific RNA transcripts bound by RBPs-of-interest.