A novel k-mer set memory (KSM) motif representation improves regulatory variant prediction

Abstract

The representation and discovery of transcription factor (TF) sequence binding specificities is critical for understanding gene regulatory networks and interpreting the impact of disease-associated noncoding genetic variants. We present a novel TF binding motif representation, the -mer set memory (KSM), which consists of a set of aligned-mers that are overrepresented at TF binding sites, and a new method called KMAC for de novo discovery of KSMs. We find that KSMs more accurately predict in vivo binding sites than position weight matrix (PWM) models and other more complex motif models across a large set of ChIP-seq experiments. Furthermore, KSMs outperform PWMs and more complex motif models in predicting in vitro binding sites. KMAC also identifies correct motifs in more experiments than five state-of-the-art motif discovery methods. In addition, KSM-derived features outperform both PWM and deep learning model derived sequence features in predicting differential regulatory activities of expression quantitative trait loci (eQTL) alleles. Finally, we have applied KMAC to 1600 ENCODE TF ChIP-seq data sets and created a public resource of KSM and PWM motifs. We expect that the KSM representation and KMAC method will be valuable in characterizing TF binding specificities and in interpreting the effects of noncoding genetic variations.

Motto: Representing Motifs in Consensus Sequences with Minimum Information Loss

“…The MEME format is supported by the majority of the motif databases ( Kulakovskiy et al 2018 ), and the MEME suite provides packages for integrative analysis and conversion from other motifs formats ( Bailey et al 2009 ). The recently proposed kmer-based motif models also support conversion to MEME format ( Fletez-Brant et al 2013 ; Ghandi et al 2014 ; Zeng et al 2016 ; Guo et al 2018 ). Our package is lightweight and open-source.…”

“…However, when in a textual interface, representing PWMs requires an by matrix, where is the number of characters (such as A, C, G, T, for nucleotides), and is the length of the motif. Recently, several studies have shown the usefulness of representing motifs using kmers ( Fletez-Brant et al 2013 ; Ghandi et al 2014 ; Zeng et al 2016 ; Guo et al 2018 ); despite the power of this representation in machine learning models, it is cumbersome to have a set of kmers to characterize a single motif. In many scenarios, motifs can be sufficiently represented by regular expressions of the consensus sequences, such as [GC][AT]GATAAG[GAC] for the GATA2 motif.…”

Prediction of regulatory motifs from human Chip-sequencing data using a deep learning framework

“…MEME (51) and DREME (52). KMAC innovatively used k -mer set memory for motif representation in order to capture the contribution of nucleotides dependency and flanking k -mers in TF–DNA binding (32). Compared with other state-of-the-art motif finding methods (e.g.…”

“…Compared with other state-of-the-art motif finding methods (e.g. HOMER (35) and ChIPMunk (53)), KMAC achieved the best performance in discovering known motifs from ChIP-seq datasets (32). Gkm-SVM was selected in the comparison as it significantly outperforms traditional kmer-SVM methods by using gapped k -mers for accurately and efficiently identifying longer motifs, which are hard to model as k -mers.…”

“…For the first time, these DNA shape features were integrated into DL models to explore how these features quantitatively contribute to TF–DNA binding, even though conserved shape patterns (or shape motifs) encoded in the human genome are still not well modeled (27). The identified motifs were evaluated using the documented motifs in JASPAR (28) and TRANSFAC (29), then compared with DeepBind (25), Basset (30), MEME-ChIP (31), KMAC (32) and gkm-SVM (33). Further analyses were conducted by integrating multiple types of biological information including TF binding domain types, chromatin accessibility, phylogenetic conservation, protein–protein interactions (PPI), etc.…”