Automated tertiary structure prediction with accurate local model quality assessment using the IntFOLD-TS method

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein structure prediction

Background:

• Accurate protein structure prediction is crucial for understanding biological function.
• Model quality assessment (QA) is a critical step in protein modeling pipelines.
• Assessing local errors in predicted structures aids experimental validation.

Purpose of the Study:

• To develop and evaluate the IntFOLD-TS method for template-based protein modeling.
• To integrate a robust QA stage into the modeling pipeline.
• To provide accurate predictions of both global and local model quality.

Main Methods:

• Generation of multiple alternate models using various sequence-structure alignment methods.
• Ranking of models using the ModFOLDclust2 quality assessment tool.
• Prediction of per-residue local errors correlating to native structure deviations.

Main Results:

• IntFOLD-TS successfully generated high-quality 3D models for numerous CASP9 targets.
• The method demonstrated high accuracy in predicting per-residue errors.
• Predicted local error scores closely matched deviations from native structures.

Conclusions:

• The IntFOLD-TS method provides reliable 3D protein models with accurate quality assessments.
• Inclusion of local error predictions enhances the practical utility of predicted structures.
• This approach can guide and inform future experimental investigations in structural biology.