PreHots

PreHots predicts protein–DNA binding energy hot spots to identify residues that significantly contribute to binding free energy at protein–DNA interfaces.

Key Features:

• Ensemble stacking classifier: PreHots uses an ensemble stacking classifier that integrates multiple machine learning classifiers to predict hot spots.
• Feature selection: The model employs 19 features selected by a sequential backward feature selection algorithm.
• Dataset construction: Two manually curated datasets containing 123 hot spots and 137 non-hot spots from 89 protein–DNA complexes were compiled from literature and databases with redundancy removal for training and validation.
• Performance metrics: In 10-fold cross-validation on the benchmark dataset it achieved sensitivity 0.813 and AUC 0.868, and on an independent test set it achieved sensitivity 0.818 and AUC 0.820.

Scientific Applications:

• Protein–DNA interaction analysis: Predicting hot spots to elucidate residues that dominate binding free energy at protein–DNA interfaces.
• Drug design: Informing the selection of interface residues relevant to targeting protein–DNA interactions.
• Genetic regulation studies: Identifying residues that may modulate regulatory protein–DNA binding.
• Molecular biology research: Supporting studies of molecular mechanisms underlying protein–DNA recognition and binding energetics.

Methodology:

An ensemble stacking classifier integrating multiple machine learning classifiers was trained using 19 features selected by sequential backward feature selection on two manually curated datasets (123 hot spots, 137 non-hot spots from 89 complexes), with performance evaluated by 10-fold cross-validation and an independent test set.