Predicting DNA-binding proteins and binding residues by complex structure prediction and application to human...

Area of Science:

• Structural biology
• Computational biology
• Bioinformatics

Background:

• Rapid advancements in sequencing technologies yield vast numbers of protein sequences.
• Determining protein function, particularly DNA-binding capabilities, is crucial for biological understanding.
• Existing methods often provide low-resolution predictions, lacking structural detail.

Purpose of the Study:

• To develop a reliable computational technique for predicting protein DNA-binding function at the protein-DNA complex structure level.
• To improve upon existing two-state DNA-binding prediction methods by incorporating structural information.
• To identify novel DNA-binding proteins within proteomes.

Main Methods:

• Utilized template-based structure prediction (HHblits) to predict protein-DNA complex structures.
• Employed a knowledge-based energy function (Distance-scaled finite ideal-gas reference state) for binding affinity prediction.
• Validated the method using leave-one-out cross-validation on known DNA-binding and non-binding protein domains.

Main Results:

• Achieved a Matthews correlation coefficient (MCC) of 0.77 with high precision (94%) and sensitivity (65%) in cross-validation.
• Demonstrated 51% sensitivity on newly determined DNA-binding protein structures and 56% on the human proteome.
• Successfully predicted DNA-binding residues and identified over 300 potential novel DNA-binding proteins in the human proteome.

Conclusions:

• The SPOT-Seq (DNA) method offers a highly reliable approach for predicting protein DNA-binding function using structural information.
• The method accurately predicts DNA-binding proteins and residues, aiding in functional annotation and discovery.
• SPOT-Seq (DNA) has significant implications for identifying novel DNA-binding proteins and understanding their roles in biological processes.