Estimating the Quality of 3D Protein Models Using the ModFOLD7 Server

Area of Science:

• Computational biology
• Structural bioinformatics

Background:

• Accurate assessment of 3D protein models is crucial in structural biology.
• Existing methods for Estimates of Model Accuracy (EMA) have limitations in scoring and ranking predicted protein structures.

Purpose of the Study:

• To present ModFOLD7, an upgraded EMA resource for assessing protein model accuracy.
• To improve the accuracy and consistency of scoring and ranking predicted protein models.

Main Methods:

• Integration of pioneering pure-single- and quasi-single-model approaches into the ModFOLD resource.
• Development of three distinct global score estimation options: ModFOLD7_rank, ModFOLD7_cor, and ModFOLD7 global.

Main Results:

• ModFOLD7 demonstrates enhanced accuracy in scoring and ranking predicted protein models compared to previous EMA methods.
• Independent blind testing by CASP13 assessors ranked ModFOLD7 among the top EMA methods.
• Continuous evaluation in the CAMEO project shows significant improvement over earlier versions.

Conclusions:

• ModFOLD7 represents a significant advancement in computational tools for protein structure prediction accuracy assessment.
• The server offers flexible scoring options catering to diverse user needs in model evaluation.
• ModFOLD7 is freely available, supporting the broader scientific community in structural bioinformatics research.