F-Seq2: improving the feature density based peak caller with dynamic statistics.

Area of Science:

• Genomics and Epigenomics
• Bioinformatics
• Computational Biology

Background:

• High-throughput sequencing (HTS) technologies capture genome-wide genomic and epigenomic features.
• Peak calling algorithms identify functional genomic regions, such as open chromatin and transcription factor binding sites, from HTS data.
• F-Seq1, a previous peak caller, showed high sensitivity for DNase-seq data but lacked support for control datasets and robust test statistic reporting, limiting bias assessment.

Purpose of the Study:

• To address the limitations of F-Seq1 by developing an improved peak calling algorithm.
• To enhance the accuracy and reliability of peak calling for various HTS epigenomic datasets.
• To provide a tool that accounts for local biases and accurately ranks predicted peaks by confidence.

Main Methods:

• F-Seq2 was developed, integrating kernel density estimation and a dynamic 'continuous' Poisson test.
• The algorithm accounts for local biases in read distributions during peak prediction.
• Test statistics are calculated for individual candidate summits, enabling assessment of irreproducible discovery rate (IDR).

Main Results:

• F-Seq2 demonstrates improved performance for ATAC-seq and ChIP-seq datasets compared to existing peak callers.
• The new method effectively accounts for local biases in background distributions.
• F-Seq2 provides confidence rankings for predicted peaks, suitable for downstream analyses like IDR.

Conclusions:

• F-Seq2 represents a significant advancement in peak calling for epigenomic studies.
• The tool offers enhanced precision and recall, outperforming current standards, including those used by the ENCODE Consortium.
• F-Seq2 facilitates more reliable identification and analysis of functional genomic elements from HTS data.