Prediction of protein assemblies, the next frontier: The CASP14-CAPRI experiment

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• The Critical Assessment of PRedicted Interactions (CAPRI) challenge assesses protein assembly prediction accuracy.
• CAPRI Round 50 was the fourth joint challenge with the Critical Assessment of Structure Prediction (CASP).
• This round included 12 targets: 6 dimers, 3 trimers, and 3 higher-order oligomers, with varying difficulty based on template availability.

Purpose of the Study:

• To evaluate the performance of computational methods in predicting protein complex structures.
• To assess the accuracy of protein assembly predictions in the fourth joint CASP-CAPRI challenge.
• To identify advances in protein assembly prediction methodologies.

Main Methods:

• Twenty-five CAPRI groups, including eight automatic servers, submitted approximately 1250 models per target.
• Twenty groups, including six servers, participated in the scoring challenge, submitting around 190 models per target.
• Model accuracy was assessed using standard CAPRI criteria and a weighted scoring scheme evaluating top-ranking models.

Main Results:

• Top-performing groups submitted acceptable or better models for 70-75% of targets, an increase from previous rounds, but with lower high-accuracy rates.
• Scorer groups demonstrated improved performance, with more groups submitting correct models for 70-80% of targets or achieving high accuracy.
• Automatic servers generally underperformed, with MDOCKPP and LZERD being notable exceptions, matching human group performance.

Conclusions:

• Protein assembly prediction has seen methodological advances, though achieving high accuracy remains challenging.
• Scoring methods show improved effectiveness in evaluating protein complex predictions.
• While general server performance lagged, specific servers achieved competitive results, indicating progress in automated structure prediction.