Evaluation of disorder predictions in CASP9

Area of Science:

• Protein structure and function
• Bioinformatics and computational biology
• Intrinsically disordered proteins

Background:

• Intrinsically disordered proteins (IDPs) lack stable 3D structure but are vital for various cellular functions.
• Identifying disordered regions is essential for understanding protein mechanisms.
• The Critical Assessment of protein Structure Prediction (CASP) initiative evaluates computational prediction methods.

Purpose of the Study:

• To evaluate the performance of protein disorder prediction methods submitted to CASP9.
• To compare CASP9 results with previous CASP assessments.
• To analyze the impact of disorder segment length on prediction accuracy.

Main Methods:

• Utilized standard CASP evaluation metrics: balanced accuracy and Matthews correlation coefficient for binary classification.
• Employed the area under the receiver operating characteristic curve for probability-based predictions.
• Analyzed prediction performance across different minimum disorder segment lengths.

Main Results:

• CASP9 disorder prediction methods demonstrated a slight improvement compared to CASP7.
• Performance of CASP9 methods did not surpass that of CASP8 methods.
• Prediction accuracy generally decreased as the minimum length of the disorder segment increased.

Conclusions:

• Computational methods for predicting protein disorder continue to evolve, with incremental gains observed.
• The challenge of accurately predicting longer intrinsically disordered regions remains.
• Continued assessment through initiatives like CASP is vital for advancing the field of protein structure and function prediction.