iDHS-DMCAC: identifying DNase I hypersensitive sites with balanced dinucleotide-based detrending moving-average...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• DNase I hypersensitive sites (DHSs) are crucial regulatory DNA elements impacting biomedical research and drug discovery.
• Experimental methods for DHS identification are laborious, time-consuming, and often inaccurate.
• The increasing volume of genomic data necessitates efficient computational approaches for DHS identification.

Purpose of the Study:

• To develop a cost-effective computational model for identifying DHSs.
• To introduce a novel statistical feature extraction method for DHS prediction.
• To provide a high-throughput tool for analyzing genomic regulatory elements.

Main Methods:

• A statistical feature extraction model, iDHS-DMCAC, was developed using the detrended moving-average cross-correlation (DMCA) coefficient descriptor.
• A 105-dimensional feature vector was constructed based on a dinucleotide property matrix derived from 15 DNA dinucleotide properties.
• The model employed over-sampling techniques and support vector machine algorithms for classification on imbalanced datasets.

Main Results:

• The iDHS-DMCAC model demonstrated superior accuracy and stability compared to existing methods in rigorous cross-validations.
• The predictor achieved high performance on benchmark datasets, highlighting its effectiveness.
• The developed model provides a robust computational approach for DHS identification.

Conclusions:

• iDHS-DMCAC is a highly accurate and stable computational tool for identifying DNase I hypersensitive sites.
• This model offers a valuable high-throughput alternative or complement to traditional experimental methods.
• The freely available datasets and source code facilitate further research and application in genomics.