EL_PSSM-RT

EL_PSSM-RT predicts DNA-binding residues by combining Position Specific Score Matrix Relation Transformation (PSSM-RT) residue encoding with ensemble learning to improve identification of protein–DNA recognition sites.

Key Features:

• PSSM-RT residue encoding: Encodes residues using Position Specific Score Matrix Relation Transformation to capture evolutionary information relationships and pairwise residue relationships.
• Ensemble learning classifier (EL_PSSM-RT): Integrates multiple classifiers to improve prediction accuracy for DNA-binding residues.
• Imbalance handling: Addresses the imbalance between binding and non-binding residues in datasets to enhance robustness.
• Benchmark evaluation: Evaluated using five-fold cross-validation on PDNA-62, PDNA-224, TS-72, and TS-61.
• Reported performance gains: Demonstrates improvements in Matthews Correlation Coefficient (MCC) by 0.02–0.07, sensitivity (ST) by 4.18%–21.47%, and Area Under the Curve (AUC) by 0.013–0.131.

Scientific Applications:

• Performance evaluation: Five-fold cross-validation on PDNA-62, PDNA-224, TS-72, and TS-61 shows EL_PSSM-RT outperforms existing predictors by the reported MCC, ST, and AUC ranges.
• Validation of evolutionary information: Provides a validated approach that highlights the importance of evolutionary information relationships between residues for DNA-binding site prediction.
• Protein-DNA interaction studies: Supports analysis of protein–DNA recognition, gene regulation, and transcriptional control by predicting DNA-binding residues.
• Drug discovery and development: Identifies critical DNA-binding residues that can inform therapeutic target identification.

Methodology:

Position Specific Score Matrix Relation Transformation (PSSM-RT) residue encoding; ensemble learning classifier EL_PSSM-RT to combine multiple classifiers and handle class imbalance; evaluation via five-fold cross-validation on PDNA-62, PDNA-224, TS-72, and TS-61.