Sphinx: merging knowledge-based and ab initio approaches to improve protein loop prediction.

Area of Science:

• Structural biology
• Computational biology
• Bioinformatics

Background:

• Protein loops are crucial for function but difficult to model accurately due to irregular structures.
• Current loop modeling methods are primarily knowledge-based (fragment searching) or ab initio (computational generation).
• Existing knowledge-based methods are limited by using only same-length fragments, ignoring conformational similarities in loops of different lengths.

Purpose of the Study:

• To develop a novel method, Sphinx, for improved protein loop structure prediction.
• To integrate ab initio techniques with structural information from loops of varying lengths.
• To enhance the accuracy and scope of protein loop modeling.

Main Methods:

• Sphinx combines ab initio loop generation with the use of structural information from loops of different lengths.
• The method leverages potential extra structural insights from non-identical loop lengths.
• Sphinx aims to overcome limitations of traditional knowledge-based and ab initio approaches.

Main Results:

• Sphinx achieves high-accuracy predictions and generates decoy sets rich in near-native conformations.
• The method outperforms the ab initio algorithm it is based upon.
• Sphinx successfully predicts antibody H3 structures, surpassing RosettaAntibody in accuracy and speed.

Conclusions:

• Sphinx offers a significant advancement in protein loop modeling.
• The method provides accurate predictions for all targets, including challenging antibody H3 regions.
• Sphinx represents a more versatile and effective tool for protein structure prediction.