SplitSeek-Pro: accurate prediction of splittable sites on protein structures

Area of Science:

• Protein engineering
• Computational biology
• Structural biology

Background:

• Protein architecture understanding is crucial for engineering functional proteins.
• Predicting structural tolerance to remodeling is key for protein design.
• Splittability is a critical property for protein engineering tasks.

Purpose of the Study:

• To develop a deep learning model, SplitSeek-Pro, for evaluating amino acid-level splittability in folded proteins.
• To provide a tool for protein engineering tasks like circular permutation and split reconstitution.
• To identify characteristic protein segments functioning as cohesive fragments.

Main Methods:

• Integration of primary amino acid sequences with 3D structural features.
• A two-stage deep learning training process: large-scale pre-training and high-quality fine-tuning.
• Residue-resolution prediction of protein splittability.

Main Results:

• SplitSeek-Pro demonstrates superior predictive power compared to existing methods.
• Experimental validation on three distinct proteins confirmed the model's effectiveness.
• Identification of characteristic protein segments analogous to super-secondary structural motifs.

Conclusions:

• SplitSeek-Pro is a robust tool for rational protein engineering.
• The model offers insights into the fundamental structural building blocks of protein folding.
• An automated web server is available for community access.