Multiple conformational states assembly of multidomain proteins using evolutionary algorithm based on structural...

Area of Science:

• Structural Biology
• Computational Biology
• Biophysics

Background:

• AlphaFold2 revolutionized single-domain protein structure prediction.
• Accurate modeling of multidomain proteins, especially their multiple conformational states, remains a significant challenge.

Purpose of the Study:

• To develop an advanced method, M-SADA, for assembling multidomain proteins with multiple conformational states.
• To improve the accuracy and completeness of protein structure modeling for complex biological systems.

Main Methods:

• Developed M-SADA, a multidomain protein assembly method utilizing a multiple population-based evolutionary algorithm.
• Integrated homologous and analogous templates with deep learning-predicted inter-domain distances to guide energy function construction.
• Created benchmark datasets for evaluating performance on both multiple and single conformational states.

Main Results:

• M-SADA significantly outperforms AlphaFold2 in modeling multiple conformational states of multidomain proteins (40.3% achieved TM-score > 0.90 for distinct states).
• M-SADA shows a 5.2% higher average TM-score (0.913 vs. 0.868) than AlphaFold2 for single conformational state modeling on a large benchmark dataset.
• M-SADA successfully predicted novel domain arrangements for CASP15 multidomain targets when individual domain structures were accurate.

Conclusions:

• M-SADA represents a significant advancement in computational protein structure modeling, particularly for multidomain proteins with dynamic conformational ensembles.
• The method offers improved accuracy for both single and multiple conformational state predictions, addressing limitations of existing tools like AlphaFold2.
• M-SADA has the potential to accelerate research in areas requiring precise understanding of protein dynamics and function.