Evaluation of predictions in the CASP10 model refinement category

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• Assessing the state-of-the-art in protein model refinement is crucial for advancing structural biology.
• Previous experiments (e.g., CASP9) highlighted limitations in current protein model refinement techniques.

Purpose of the Study:

• To evaluate the performance of various methods in improving the accuracy of initial protein models.
• To identify successful strategies and remaining challenges in protein structure refinement.

Main Methods:

• Participants were tasked with refining 27 initial protein models.
• A range of complementary evaluation measures were employed to assess model accuracy.
• Performance was compared against a naïve (null) method and previous benchmarks (CASP9).

Main Results:

• Five groups demonstrated performance superior to the naïve method, with three showing statistically significant improvements.
• Leading groups effectively refined both backbone and side chain positioning.
• The top-ranked group achieved backbone conformation improvement in nearly 90% of targets, a novel success rate.

Conclusions:

• Significant progress has been made in protein model refinement, surpassing previous benchmarks.
• While improvements are evident, challenges remain in consistently enhancing model quality and achieving substantial accuracy gains.
• Further research is needed to address limitations and ensure successful refinement attempts are fully recognized.