BioTile, A Perl based tool for the identification of differentially enriched regions in tiling microarray data

Jerry Guintivano,Zachary A Kaminsky,Todd D Gould,Kellie Lk Tamashiro,Michal Arad

doi:10.1186/1471-2105-14-76

Abstract

BackgroundGenome-wide tiling array experiments are increasingly used for the analysis of DNA methylation. Because DNA methylation patterns are tissue and cell type specific, the detection of differentially methylated regions (DMRs) with small effect size is a necessary feature of tiling microarray ‘peak’ finding algorithms, as cellular heterogeneity within a studied tissue may lead to a dilution of the phenotypically relevant effects. Additionally, the ability to detect short length DMRs is necessary as biologically relevant signal may occur in focused regions throughout the genome.ResultsWe present a free open-source Perl application, Binding Intensity Only Tile array analysis or “BioTile”, for the identification of differentially enriched regions (DERs) in tiling array data. The application of BioTile to non-smoothed data allows for the identification of shorter length and smaller effect-size DERs, while correcting for probe specific variation by inversely weighting on probe variance through a permutation corrected meta-analysis procedure employed at identified regions. BioTile exhibits higher power to identify significant DERs of low effect size and across shorter genomic stretches as compared to other peak finding algorithms, while not sacrificing power to detect longer DERs.ConclusionBioTile represents an easy to use analysis option applicable to multiple microarray platforms, allowing for its integration into the analysis workflow of array data analysis.

Highlights

Genome-wide tiling array experiments are increasingly used for the analysis of DNA methylation
In order to minimize the identification of false positives and maximize differentially enriched regions (DERs) peak identification, BioTile employs the use of a permutation corrected meta-analysis step capable of detecting peaks comprised of as few as three adjacent probes
As BioTile identifies DERs by identifying regions where the pair wise difference between the two groups is consistently above or below zero for a stretch longer than 3 probes, if false positive signals are located within regions where background levels are in the same direction, they will be included in the list of DERs to evaluate statistically

Summary

Results

We present a free open-source Perl application, Binding Intensity Only Tile array analysis or “BioTile”, for the identification of differentially enriched regions (DERs) in tiling array data. The application of BioTile to non-smoothed data allows for the identification of shorter length and smaller effect-size DERs, while correcting for probe specific variation by inversely weighting on probe variance through a permutation corrected meta-analysis procedure employed at identified regions. BioTile exhibits higher power to identify significant DERs of low effect size and across shorter genomic stretches as compared to other peak finding algorithms, while not sacrificing power to detect longer DERs

Conclusion

Background

Results and discussion

27. Cochran WG