Assessing deep learning algorithms incis-regulatory motif finding based on genomic sequencing data

Abstract

Cis-regulatory motif finding is a crucial step in the detection of gene regulatory mechanisms using genomic data. Deep learning (DL) models have been utilized to de-novoly identify motifs, and have been proven to outperform traditional methods. By 2020, twenty DL models have been developed to identify DNA and RNA motifs with diverse framework designs and implementation styles. Hence, it is beneficial to systematically compare their performances, which can facilitate researchers in selecting the appropriate tools for their motif analyses. Here, we carried out an in-depth assessment of the 20 models utilizing 1,043 genomic sequencing datasets, including 690 ENCODE ChIP-Seq, 126 cancer ChIP-Seq, 172 single-cell cleavages under targets and release using a nuclease, and 55 RNA CLIP-Seq. Four metrics were designed and investigated, including the accuracy of motif finding, the performance of DNA/RNA sequence classification, algorithm scalability, and tool usability. The assessment results demonstrated the high complementarity of the existing models, and it was determined that the most suitable model should primarily depend on the data size and type as well as the model outputs. A webserver was developed to allow efficient access of the identified motifs and effective utilization of high-performing DL models.

GraphPred: An approach to predict multiple DNA motifs from ATAC-seq data using graph neural network and coexisting probability

“…For each ATAC-seq experiment, we found motifs by the above way and retained motifs that ICs at least three positions are larger than 1. The found motifs are compared to the HOCOMOCO v11 database via the TOMTOM v5.1.0 tool [2, 41]. The matched motifs are significant if their p-values are less than 0.05.…”

“…ATAC-seq obtains more kinds of transcription factor binding regions than Chromatin immunoprecipitation sequencing (ChIP-seq) and DNase I hypersensitive sites sequencing (DNase-seq). The sites of Transcription factors (TFs) binding to DNA sequences are called transcription factor binding sites (TFBSs), which are usually short DNA sequences [2]. The aligned TFBSs of the same TF is the regulatory motif, which can be represented by the position weight matrix (PWM) [3, 4].…”