Terminitor: Cleavage Site Prediction Using Deep Learning Models

Abstract

As a widespread RNA processing machinery, alternative polyadenylation plays a crucial role in gene regulation. To help decipher its underlying mechanism and understand its impact, it is desirable to comprehensively profile 3'-untranslated region cleavage and associated polyadenylation sites. State-of-the-art polyadenylation site detection tools are influenced either by library preparation or manually selected features. Here we present Termin(A) n tor, a deep neural network-based profiling pipeline to predict polyadenylation sites from RNA-seq data. We show how Termin(A) n tor outperforms competing tools in sensitivity and precision on experimental transcriptome sequence data. We also demonstrate applications of Termin(A) n tor with both short-read and long-read sequencing technologies.

A Survey on Methods for Predicting Polyadenylation Sites from DNA Sequences, Bulk RNA-seq, and Single-cell RNA-seq

“…IntMAP [97] leverages one unified ML framework to combine the information from RNA-seq and 3′ seq to quantify different 3′ UTR isoforms using a global optimization strategy. Terminitor [98] is based on a deep . CC-BY-NC-ND 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.…”

“…In recent years, some newly emerging methods employ traditional ML or DL model to identify pAs from RNA-seq, including TECtools [96], IntMAP [97], Terminitor [98], and Aptardi [99]. TECtools [96] first identifies terminal exons and transcript isoforms ending at known intronic pAs.…”

“…IntMAP [97] leverages one unified ML framework to combine the information from RNA-seq and 3′ seq to quantify different 3′ UTR isoforms using a global optimization strategy. Terminitor [98] is based on a deep neural network for three-label classification problem, which can determine whether an input sequence contains a pA with poly(A) signal, a site without poly(A) signal, or non-pA.…”

“…IntMAP [97] leverages one unified ML framework to combine the information from RNA-seq and 3’ seq to quantify different 3’ UTR isoforms using a global optimization strategy. Terminitor [98] is based on a deep neural network for three-label classification problem, which can determine whether an input sequence contains a pA with poly(A) signal, a site without poly(A) signal, or non-pA. Aptardi [99] is a multi-omics approach based on bidirectional long short-term memory recurrent neural network (biLSTM), which predicts pAs by leveraging DNA sequences, RNA-seq, and the predilection of transcriptome assemblers.…”

Deep Learning in Protein Structural Modeling and Design

“…The property to predict can either be a protein-level property, such as a classification as an enzyme or non-enzyme, 59 or a residue-level property, such as the sites or motifs of phosphorylation (DeepPho) 60 and polyadenylation (Terminitor). 61 The challenging part here and in the following models is how to represent the protein. Representation refers to the encoding of a protein that serves as an input for prediction tasks or the output for generation tasks.…”