On predicting foldability of a protein from its sequence.

Area of Science:

• Computational Biology
• Protein Bioinformatics
• Structural Bioinformatics

Background:

• Understanding protein folding is crucial for deciphering biological function and disease mechanisms.
• Distinguishing sequences that fold from those that do not is a fundamental challenge in protein science.
• Intrinsically disordered proteins (IDPs) represent a distinct class with unique sequence characteristics.

Purpose of the Study:

• To identify specific properties of amino acid sequences that differentiate folding proteins from random sequences and intrinsically disordered proteins.
• To develop and evaluate predictive metrics for protein foldability based on sequence properties.

Main Methods:

• Comparison of amino acid sequence properties, including secondary structure propensities (helix, sheet), adjacency correlations, and directionality correlations.
• Analysis of propensities for all possible triplets and quadruplets of amino acids.
• Development of 'folding scores' based on observed sequence property differences and assessment of their predictive accuracy for foldability.

Main Results:

• Minor differences were found between folded and random sequences in adjacency and directional correlations.
• Significant differences emerged in triplet and, notably, quadruplet propensities between folded and non-folded sequences.
• Protein foldability prediction accuracy was highest when utilizing quadruplet propensity statistics.

Conclusions:

• Amino acid sequence quadruplet propensities are powerful indicators for predicting protein foldability.
• Sequence-based statistical analysis, particularly at the quadruplet level, offers a viable approach to discriminate folding proteins.
• These findings contribute to a deeper understanding of the sequence determinants of protein structure formation.