Revisiting the prediction of protein function at CASP6

Area of Science:

• Computational biology
• Structural genomics
• Bioinformatics

Background:

• Predicting protein function from sequence or structure is vital for utilizing vast genomic data.
• The Critical Assessment of Techniques for Protein Structure Prediction (CASP) experiment introduced a protein function prediction category to assess current methods.
• Assessing prediction accuracy was challenging due to unavailable experimental functions and blind, unverified predictions.

Purpose of the Study:

• To analyze the performance of protein function prediction methods assessed in CASP6.
• To investigate the convergence of predictions from methods with different underlying rationales.
• To propose a classification system for prediction method types and redundancy.

Main Methods:

• Collected information on methods used by CASP6 protein function predictors.
• Re-evaluated CASP6 results by analyzing convergent predictions from diverse methodologies.
• Classified prediction methods and analyzed cases where target protein functions became available.

Main Results:

• Predictions based on a consensus of different methods achieved up to 80% accuracy.
• Analysis revealed instances where convergent predictions were obtained using distinct computational approaches.
• Re-analyzed CASP6 predictions, considering method convergence, provide significant information for protein function research.

Conclusions:

• A consensus approach using diverse protein function prediction methods enhances prediction accuracy.
• The study provides a framework for classifying and understanding the redundancy of prediction methods.
• Re-analyzed CASP6 predictions offer valuable functional insights for researchers, especially when considering method convergence.