Topological Machine Learning for Protein-Nucleic Acid Binding Affinity Changes Upon Mutation.

Area of Science:

• Computational Biology
• Biophysics
• Machine Learning

Background:

• Understanding protein-nucleic acid interactions is vital for disease research and drug development.
• Existing experimental and computational methods for predicting mutation effects on binding are limited.
• Accurate prediction of mutation-induced binding changes is a significant challenge.

Purpose of the Study:

• To develop a novel computational model for predicting the impact of protein mutations on protein-nucleic acid binding affinity.
• To integrate diverse data types including topological, physicochemical, and sequence-based features for robust interaction modeling.

Main Methods:

• A topological machine learning model (TopoML) was developed, incorporating persistent Laplacian from topological data analysis.
• Multi-perspective features were used: physicochemical properties, topological structures, and protein Transformer embeddings.
• The model was validated on protein-DNA and protein-RNA datasets containing single-point amino acid mutations.

Main Results:

• The TopoML model demonstrated superior performance compared to state-of-the-art methods.
• Accurate prediction of mutation-induced binding affinity changes for both protein-DNA and protein-RNA complexes was achieved.
• The integrative framework effectively captured complex binding interaction representations.

Conclusions:

• The proposed TopoML model offers a significant advancement in predicting mutation effects on protein-nucleic acid binding.
• This method can aid in unraveling disease mechanisms and accelerating the development of targeted therapies.
• TopoML provides a more accurate and efficient computational tool for studying protein-nucleic acid interactions.