MotifbreakR v2: extended capability and database integration

Summary

MotifbreakR is a software tool that scans genetic variants against position weight matrices of transcription factors (TF) to determine the potential for the disruption of TF binding at the site of the variant. It leverages the Bioconductor suite of software packages and annotations to operate across a diverse array of genomes and motif databases. Initially developed to interrogate the effect of single nucleotide variants (common and rare SNVs) on potential TF binding sites, in motifbreakR v2, we have updated the functionality. New features include the ability to query other types of more complex genetic variants, such as short insertions and deletions (indels). This function allows modeling a more extensive array of variants that may have more significant effects on TF binding. Additionally, while TF binding is based partly on sequence preference, predictions of TF binding based on sequence preference alone can indicate many more potential binding events than observed. Adding information from DNA-binding sequencing datasets lends confidence to motif disruption prediction by demonstrating TF binding in cell lines and tissue types. Therefore, motifbreakR implements querying the ReMap2022 database for evidence that a TF matching the disrupted motif binds over the disrupting variant. Finally, in motifbreakR, in addition to the existing interface, we have implemented an R/Shiny graphical user interface to simplify and enhance access to researchers with different skill sets.

1. Introduction

Prediction of the likely consequences of variants on transcription factor (TF) binding is extremely valuable for hypothesis generation in human genetics and disease research and in the study of regulatory genomics broadly [1–5]. Tools and workflows have become increasingly sophisticated using machine learning and massively parallel reporter assays [6–9]. However, it remains practical to maintain software libraries that can generate predictions based on position weight matrix (PWM) based matching analysis as a first-pass hypothesis generator or part of a larger bioinformatics workflow. MotifbreakR remains relevant, having been used in many studies of human disease, from neurodegenerative disorders to COVID-19 research [9–59]. Additionally, it has been used in a variety of studies of model organisms, including yeast, mice, pigs, and blind cavefish [60–64]. It has even been used in paleoanthropology [65] and evolutionary genomics [66].

We originally published motifbreakR for variant analysis of genome-wide association studies (GWAS) [67]. What motifbreakR does well, which has led to its continued use across fields, is enable a standardized analysis of any genome curated by Bioconductor, using any available PWM library and any conceivable format for input variants, including BED format, VCFs, lists of rsIDs, and even “custom” variants specified in a flexible BED derived format. However, the original motifbreakR did not accept other types of more complex genetic variants, such as short insertions and deletions (indels) or variants with more than one alternative allele that constitute a significant fraction of variation: ~17% of variants found in whole-genome sequencing in gnomAD and ~23% of common variants (global MAF > 0.01) annotated in dbSNP. In addition, we found that incorporating published ChIP-seq data greatly enhances the impact of individual findings. Here, we added the ability to query indels, automated the complementary identification of cognate TF ChIP-seq peaks, and created a graphical user interface to make all these pipelines accessible to non-bioinformaticians.

2. Features

2.1. Insertion-Deletion Variants

We have implemented the import and analysis of indel variants to allow for querying a broader array of variations that alter transcription factor binding in the genome. Variants are scored by scanning a motif across the reference and the alternate sequence; the returned score is the highest-scoring match (or, equivalently, the match with the lowest p-value) in the whole sequence. The effect size is the difference between the best match on the reference allele and the alternate allele. The biggest challenge in implementation is defining a coherent coordinate system for specifying the position of the matching motif. In an example where an insertion is longer than the sequence of the queried PWM, the motif could be spliced and thus destroyed by the insertion. Alternatively, it creates the motif such that PWM overlaps the beginning, is entirely contained within, or overlaps the end of the insertion. In the instances where the motif is created, the position cannot be described by coordinates on the reference genome. MotifbreakR defines the coordinates of a motif relative to the edges of the indel. The first number is the number of bases upstream (negative) or downstream (positive) of the start of the variant describing where the motif starts. The second number is the number of bases upstream or downstream of the end of the variant, indicating where the motif ends (Figure 1A).

Figure 1.

A. Here, we represent four distinct ways that an indel can interrupt a motif and the coordinates that describe the position of the motif relative to the indel. A dark gray box highlights the consensus sequence of the motif, and the light grey box indicates the position of the indel.

B. A genome browser shot showing the export of motifbreakR results colored by the effect size of the variant. A darker color indicates a stronger effect. The remaining tracks illustrate native tracks available on the UCSC genome browser, “ReMap density” (the density of ReMap2022 results), and a representative sample of individual ChIP-Seq binding peaks for CTCF in colon samples from the same browser track. The motifbreakR results indicate the potential for a CTCF motif to be disrupted by the alternate allele of the variant, and ReMap2022 indicates evidence that CTCF can be found binding to the site of the variant.

2.2. Querying of Transcription Factor Binding Database

While computational prediction of differential transcription factor binding potential based on sequence preference is the core of motifbreakR, grounding the analysis in observed transcription factor binding can improve the prioritization of results. To this end, we incorporate querying the ReMap2022 database of DNA-binding sequencing datasets to determine in-vitro evidence for transcription factor binding at the location of a motifbreakR result. In keeping with the existing flexibility of the motifbreakR package, with regards to species being investigated, one may query transcription factor binding data in Homo sapiens (hg38, or hg19 as liftover), Mus musculus (mm10, or mm39 as liftover), Drosophila melanogaster (dm6), and Arabidopsis thaliana (TAIR10).

The new functions empower the user to build a local database and query a motifbreakR result. Once built, motifbreakR can rapidly annotate its results with ReMap sourced transcription factor peaks corresponding to motif/transcription factor relationships provided by the constituent public MotifDb [68] sources. The user may optionally query an expanded motif/transcription factor relationship encompassing the entire potential transcription factor family as implemented by MotifDb based on TFClass [68,69]. Though not comprehensive, including the experiment biotype (cell lines and tissue types) from ReMap may also be helpful for hypothesis generation. Figure 1B shows an example of a variant rs143969848 that breaks a CTCF motif centered on a ChIP-seq peak in multiple cell lines.

2.3. Graphical User Interface

To make motifbreakR accessible to bench scientists, researchers with various skill sets, and others wishing to explore its capabilities on the web, we developed an R/Shiny-based graphical user interface (GUI) that facilitates all functions of the underlying motifbreakR package. In a workflow mirroring the code-based method, the user specifies individual rsIDs or “custom” variants and performs downstream tasks via pulldown menus and radio buttons. Like the R package version, users may upload VCFs or lists of SNVs in BED format. Any figures generated are downloadable/saved to the local environment. Finally, to promote reproducible science, analysis performed with the GUI will be output as code in R markdown format for publication.

2.4. Useful exports

MotifbreakR now exports results in various tabular formats compatible with database programs, including Excel or SQL. In addition, motifbreakR can export a BED file of user-defined subsets of top matches for display in browsers such as UCSC genome browser [70], WashU Epigenome browsers [71,72], or IGV [73,74]. Optionally, matches can be color-coded by motif quality (p-value-oriented) or disruptiveness (score-oriented).

3. Conclusion

We have updated motifbreakR to version 2 to include several new useful features, most notably indels, a new analysis pipeline to reference published ChIP-seq experiments that match motifbreakR predictions, and a GUI to promote accessibility and reproducibility.