CAP: Commutative algebra prediction of protein-nucleic acid binding affinities

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Predicting protein-nucleic acid binding affinity is crucial for understanding genomic processes.
• Current methods often lack a balance between accuracy, interpretability, and computational efficiency.

Purpose of the Study:

• Introduce the Commutative Algebra Prediction (CAP) framework for accurate protein-nucleic acid binding affinity prediction.
• Develop a method that is computationally efficient and interpretable.

Main Methods:

• Coupling persistent Stanley-Reisner theory with advanced sequence embedding.
• Encoding proteins using transformer-learned embeddings capturing long-range evolutionary context.
• Representing DNA/RNA with k-mer algebra embeddings derived from persistent facet ideals for nucleotide geometry.

Main Results:

• CAP surpasses the SVSBI protein-nucleic acid benchmark.
• Demonstrated reasonable performance on new protein-RNA and protein-nucleic acid datasets.
• CAP generalizes to any protein-nucleic acid pair using only primary sequences.

Conclusions:

• CAP offers a novel, efficient, and accurate approach for predicting protein-nucleic acid binding affinity.
• Enables genome-scale analyses without 3D structural data.
• Promises accelerated virtual screening for drug discovery and protein engineering.