AlphaFold2, SPINE-X, and Seder on Four Hard CASP Targets.

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein structure prediction

Background:

• AlphaFold2 has revolutionized protein tertiary structure prediction.
• CASP15 (Critical Assessment of protein Structure Prediction) provides a benchmark for evaluating prediction methods.
• Assessing prediction accuracy on challenging cases is crucial for identifying areas for improvement.

Purpose of the Study:

• To evaluate the performance of AlphaFold2, SPINE-X, and Seder on specific low-accuracy prediction cases from CASP15.
• To compare the accuracy of these tools in predicting secondary structure (SS) and solvent accessible surface area (ASA).
• To analyze the performance in selecting the best tertiary structure models and identify limitations of AlphaFold2.

Main Methods:

• Analysis of four specific low-accuracy cases from the CASP15 experiment.
• Comparative assessment of prediction results from AlphaFold2, SPINE-X, and Seder.
• Evaluation of secondary structure and solvent accessible surface area prediction accuracy.
• Comparison of tertiary structure model selection capabilities.

Main Results:

• AlphaFold2 generally outperformed SPINE-X in predicting secondary structure and solvent accessible surface area.
• SPINE-X showed better prediction for sheet and coil regions in some instances.
• AlphaFold2's performance in selecting tertiary structure models varied compared to Seder, with identical selections in two cases.
• AlphaFold2 tended to predict more compact structures than the native ones in the analyzed cases.

Conclusions:

• While AlphaFold2 represents a significant advancement, its tertiary structure prediction accuracy can still be enhanced, particularly in challenging cases.
• Comparative analysis highlights specific strengths and weaknesses of different protein structure prediction tools.
• Further development is needed to address limitations in predicting native protein structures accurately across all cases.