DNA-MP: a generalized DNA modifications predictor for multiple species based on powerful sequence encoding method.

Area of Science:

• Genomics and Bioinformatics
• Epigenetics and Molecular Biology

Background:

• Accurate prediction of DNA modifications is crucial for understanding cell differentiation, gene expression, and epigenetic regulation.
• Existing computational methods for DNA modification prediction have limitations in performance across multiple species due to ineffective sequence encoding.

Purpose of the Study:

• To develop a generalized computational approach, DNA-MP, for precise prediction of three different DNA modifications across multiple species.
• To introduce a novel encoding method, POCD-ND, for generating statistical representations of DNA sequences.

Main Methods:

• The DNA-MP approach utilizes the POCD-ND encoding method to capture position-specific nucleotide occurrence patterns.
• A deep forest classifier is employed for the prediction of DNA modifications.
• The POCD-ND encoder was evaluated against 32 existing encoding methods on 17 benchmark datasets across 12 species using 10 classifiers.

Main Results:

• The POCD-ND encoder significantly outperformed 32 other encoding methods across all tested classifiers.
• The DNA-MP predictor demonstrated superior performance compared to state-of-the-art methods, achieving an average accuracy improvement of 7% for 4mC, 1.35% for 5hmC, and 10% for 6mA.
• The proposed method shows enhanced accuracy across multiple species and DNA modification types.

Conclusions:

• The DNA-MP approach, with its POCD-ND encoder and deep forest classifier, offers a more precise and generalized solution for DNA modification prediction.
• This tool has the potential to advance research in epigenetics, gene regulation, and cell differentiation.
• The DNA-MP web application is publicly available to support the scientific community.