FREAD revisited: Accurate loop structure prediction using a database search algorithm.

Area of Science:

• Structural bioinformatics
• Computational biology
• Protein structure prediction

Background:

• Protein loops are highly variable and difficult to predict accurately.
• Traditional prediction methods include ab initio and database search approaches.
• Ab initio methods have recently been favored over database search.

Purpose of the Study:

• To re-evaluate the effectiveness of database search methods for protein loop prediction.
• To compare the performance of FREAD against leading ab initio methods.
• To assess the impact of sequence similarity on prediction accuracy.

Main Methods:

• Utilized FREAD, a database search method.
• Quantified sequence similarity using environment-specific substitution scores.
• Benchmarked FREAD against MODELLER, PLOP, and RAPPER on various loop sets, including CASP 7 and 8 models.

Main Results:

• FREAD significantly improves loop prediction accuracy when incorporating sequence similarity.
• FREAD outperformed ab initio methods for a substantial subset of loops (up to two-thirds of shorter, half of longer loops).
• For certain loops, FREAD achieved <2A RMSD, while other methods averaged >10A for length 20 loops.

Conclusions:

• Database search methods, particularly FREAD, are highly effective for protein loop prediction and their potential may be underestimated.
• FREAD offers competitive or superior accuracy compared to ab initio methods for identifiable loop subsets.
• Prediction accuracy using FREAD is independent of loop length and unaffected by overall model structure quality.