HSM6AP

HSM6AP predicts N6-methyladenosine (m^6A) methylation sites in Homo sapiens using machine-learning methods to prioritize genomic positions for studies of RNA modification.

Key Features:

• Multiple Weights and Feature Stitching: Applies sample weighting during training and stitches features to improve generalization from limited datasets.
• Max-Relevance-Max-Distance (MRMD) Feature Selection: Uses Max-Relevance-Max-Distance (MRMD) to select informative and nonredundant features for m^6A site recognition.
• Feature Matrix Fusion: Fuses single features into a comprehensive feature matrix to capture interactions and dependencies across sequence and genomic coordinate-derived features.
• Extreme Gradient Boosting (XGBoost): Trains predictive models using Extreme Gradient Boosting (XGBoost), an ensemble tree-based algorithm, with parameter adjustments to enhance performance.

Scientific Applications:

• m^6A Site Identification: Identifies putative N6-methyladenosine (m^6A) sites in Homo sapiens transcripts and genomic coordinates.
• RNA Processing and Regulation Studies: Enables study of the impacts of m^6A on RNA transcription, metabolism, splicing, and stability.
• Disease Association and Biomarker Discovery: Supports analyses linking m^6A modifications to diseases such as cancer and obesity and facilitates research on pathogenesis, diagnostics, and drug development.

Methodology:

Computational methods use sequence and genomic coordinate data, apply sample weighting, perform feature extraction with stitching and Max-Relevance-Max-Distance (MRMD) selection to build a fused feature matrix, and train models using Extreme Gradient Boosting (XGBoost) with parameter adjustments and evaluation on independent datasets.