Meta-4mCpred: A Sequence-Based Meta-Predictor for Accurate DNA 4mC Site Prediction Using Effective Feature...

Area of Science:

• Genomics and Epigenetics
• Bioinformatics and Computational Biology

Background:

• DNA N4-methylcytosine (4mC) is a critical epigenetic modification regulating gene expression, DNA replication, and cell cycle.
• Accurate identification of 4mC sites is essential for understanding its biological roles, but existing in silico methods lack sufficient accuracy and generalizability.

Purpose of the Study:

• To develop an advanced in silico approach for high-throughput and accurate identification of 4mC sites.
• To create Meta-4mCpred, the first meta-predictor designed specifically for 4mC site prediction, aiming to overcome limitations of current tools.

Main Methods:

• A feature representation learning scheme was utilized, generating 56 probabilistic features from four machine-learning algorithms and seven feature encodings.
• These features captured diverse sequence information, including compositional, physicochemical, and position-specific attributes.
• A support vector machine was employed with these probabilistic features to build the final meta-predictor.

Main Results:

• Meta-4mCpred achieved an average accuracy of 84.2% across six species during cross-validation, outperforming state-of-the-art predictors by 2%-4%.
• Independent dataset evaluation showed an average accuracy of 86%, exceeding current methods by over 4%.
• A user-friendly webserver for Meta-4mCpred is available at http://thegleelab.org/Meta-4mCpred.

Conclusions:

• Meta-4mCpred represents a significant advancement in computational epigenetics, offering improved accuracy and generalizability for 4mC site identification.
• The developed meta-predictor provides a valuable tool for researchers investigating the functional significance of 4mC modifications in various biological contexts.