Prediction of temperature factors from protein sequence

Area of Science:

• Protein structure and dynamics
• Computational biology
• Biophysics

Background:

• Protein flexibility is crucial for protein structure and function.
• Predicting residue flexibility from primary sequences can enhance structural and functional analyses.

Purpose of the Study:

• To identify local primary sequence features that predict amino acid residue B-values.
• To analyze B-value distribution in protein 3D structures.
• To develop a computational method for B-value prediction from protein sequences.

Main Methods:

• Analysis of local primary protein sequence features.
• Investigation of B-value distribution relative to protein surface distance, secondary structure, and residue type (hydrophobic/polar).
• Development of a Support Vector Regression (SVR) model for B-value prediction.

Main Results:

• Normalized B-value decreases with increasing distance from the protein surface.
• Loop residues exhibit higher B-values than residues in regular secondary structures.
• Buried and hydrophobic residues show lower B-values (higher rigidity) compared to surface and polar residues.
• The SVR model achieved a correlation coefficient of 0.47 for B-value prediction.

Conclusions:

• Protein residue B-values are influenced by local sequence features and structural context.
• The proposed SVR method offers a viable approach for predicting protein flexibility directly from primary sequences.
• This prediction capability can aid in understanding protein dynamics and function.