fcScan: a versatile tool to cluster combinations of sites using genomic coordinates

Abstract

Background Finding combinations of homotypic or heterotypic genomic sites obeying a specific grammar in DNA sequences is a frequent task in bioinformatics. A typical case corresponds to the identification of cis-regulatory modules characterized by a combination of transcription factor binding sites in a defined window size. Although previous studies identified clusters of genomic sites in species with varying genome sizes, the availability of a dedicated and versatile tool to search for such clusters is lacking. Results We present fcScan, an R/Bioconductor package to search for clusters of genomic sites based on user defined criteria including cluster size, inter-cluster distances and sites order and orientation allowing users to adapt their search criteria to specific biological questions. It supports GRanges, data frame and VCF/BED files as input and returns data in GRanges format. By performing clustering on vectorized data, fcScan is adapted to search for genomic clusters in millions of sites as input in short time and is thus ideal to scan data generated by high throughput methods including next generation sequencing. Conclusions fcScan is ideal for detecting cis-regulatory modules of transcription factor binding sites with a specific grammar as well as genomic loci enriched for mutations. The flexibility in input parameters allows users to perform searches targeting specific research questions. It is released under Artistic-2.0 License. The source code is freely available through Bioconductor (https://bioconductor.org/packages/fcScan) and GitHub (https://github.com/pkhoueiry/fcScan).

Guidelines on the performance evaluation of motif recognition methods in bioinformatics

“…If a novel method or dataset does not allow standard metrics, others may be used, provided that it is clearly explained why standard metrics are not applicable. One should avoid giving subjective assertions of performance like “Identified all 40 conserved modules reported previously” without mentioning how many other modules (false positives) were also identified, or referring to the literature as the only measure of correct predictions ( El-Kurdi et al, 2020 ). Reference to the literature is fully valid and useful, provided that comprehensive statistics are given.…”