When awaiting 'Bio' Champollion: dynamic programming regularization of the protein secondary structure predictions

Area of Science:

• Biochemistry
• Computational Biology
• Structural Biology

Background:

• Protein secondary structure prediction is crucial for understanding protein function.
• Existing prediction methods often generate unrealistic structures, with alpha-helices and beta-strands of inaccurate lengths.
• Current methods for improving realism rely on heuristic, intuitive, and ad hoc filtering or smoothing techniques.

Purpose of the Study:

• To introduce an objective regularization method for enhancing the realism of protein secondary structure predictions.
• To provide a method that can be integrated with any prediction technique yielding propensities.
• To improve the 'protein-likeness' of predicted structures while maintaining prediction accuracy.

Main Methods:

• Development of a regularization method based on dynamic programming algorithms.
• Application of the method to predicted secondary structure propensities.
• Evaluation of the method's impact on prediction realism and accuracy.

Main Results:

• The regularization method produces more realistic protein secondary structures compared to existing approaches.
• The method objectively corrects unrealistic predictions derived from propensities.
• The approach successfully improves the 'protein-likeness' of predicted structures.

Conclusions:

• The proposed regularization method offers an objective and effective way to improve protein secondary structure prediction realism.
• This method enhances the biological plausibility of predictions without compromising their overall accuracy.
• The regularization technique is versatile and compatible with various secondary structure prediction algorithms.