iDLDDG: predicting protein stability changes from missense mutations in DNA-binding proteins using integrated deep...

Area of Science:

• Genomics and Bioinformatics
• Molecular Biology
• Computational Biology

Background:

• Accurate prediction of missense mutations' impact on protein-DNA binding affinity is vital for disease mechanism research and therapeutic development.
• Existing models often fail to account for the distinct characteristics of mutations in double-stranded DNA-binding proteins (DSBs) and single-stranded DNA-binding proteins (SSBs).

Purpose of the Study:

• To develop a computational framework for accurately predicting the effects of missense mutations on protein-DNA binding affinity in both DSBs and SSBs.
• To establish a method that rigorously differentiates mutation mechanisms between DSBs and SSBs.

Main Methods:

• Constructed a comprehensive dataset from diverse sources.
• Developed iDLDDG, a deep learning framework integrating sequence-based embeddings (ESM2, ProtTrans, ESM1v) with multi-scale structural and evolutionary information.
• Employed an entropy-based algorithm to identify 181 optimal residues for modeling biophysical constraints, enhancing predictive accuracy and efficiency.

Main Results:

• iDLDDG achieved state-of-the-art performance, with a 10-fold cross-validation Pearson Correlation Coefficient (PCC) of 0.755 on the MPD276 dataset.
• Achieved a PCC of 0.632 on independent test sets covering both DSBs and SSBs, significantly outperforming existing methods.
• Demonstrated the framework's ability to differentiate mutation mechanisms between DSBs and SSBs.

Conclusions:

• iDLDDG provides a foundation for high-accuracy prediction of pathological mutations in DNA-binding proteins.
• This work establishes the first computational framework capable of rigorously distinguishing DSB and SSB mutation mechanisms.
• The enhanced predictive accuracy and computational efficiency support large-scale assessments of mutation effects in DNA-binding proteins.