pDHS-SVM: A prediction method for plant DNase I hypersensitive sites based on support vector machine.

Area of Science:

• Genomics
• Computational Biology
• Plant Science

Background:

• DNase I hypersensitive sites (DHSs) mark accessible chromatin and are key indicators of cis-regulatory elements (CREs) involved in gene regulation.
• Identifying DHSs is crucial for understanding gene expression, but tools for plant genomes are lacking.
• Developing cost-effective computational methods to identify DHSs is essential for accelerating plant genomic research.

Purpose of the Study:

• To develop pDHS-SVM, a novel computational predictor for identifying DHSs in plant genomes.
• To integrate global sequence-order and local DNA properties for accurate DHS prediction.
• To improve predictor performance by optimizing nucleotide physical-chemical properties.

Main Methods:

• Utilized reverse complement kmer and dinucleotide auto covariance for feature space construction.
• Employed Support Vector Machine (SVM) algorithm for DHS classification.
• Introduced a heuristic algorithm for selecting optimal nucleotide physical-chemical properties.

Main Results:

• pDHS-SVM achieved high prediction accuracies: 87.00% for Arabidopsis thaliana and 85.79% for rice (Oryza sativa).
• The method effectively integrates diverse DNA sequence features.
• An optimized subset of properties enhanced predictor performance.

Conclusions:

• pDHS-SVM is an effective and accurate computational tool for identifying plant DHSs.
• The predictor serves as a valuable complement for discovering plant cis-regulatory elements (CREs).
• The pDHS-SVM tool is publicly available as open-source code.