Prodepth

Prodepth predicts residue depth (RD) from protein sequences to determine the burial status of amino acid residues for structural bioinformatics analyses and as a complement to accessible surface area (ASA).

Key Features:

• Residue Depth Prediction: Predicts residue depth (RD) directly from protein sequence information to infer burial status of amino acids.
• Complement to ASA: Provides RD as a complementary measure to accessible surface area (ASA) for assessing residue positioning.
• Algorithm: Uses support vector regression (SVR) to model the relationship between sequence and RD.
• Sequence Encoding Schemes: Evaluated eight sequence encoding schemes capturing local and global sequence characteristics.
• Performance Metrics: Reported correlation coefficient (CC) of 0.71 and root mean square error (RMSE) of 1.74 between observed and predicted RD.
• Evaluation: Performance was assessed using 5-fold cross-validation.
• Determinants of RD: Indicates that local sequence environments are the primary determinants of residue depth while global features have marginal impact.

Scientific Applications:

• Identification of Functionally Important Residues: Uses predicted RD to help pinpoint residues that are likely critical for protein function.
• Folding Nucleus and Active Sites: Facilitates identification of folding nucleus regions and enzymatic active sites via depth information.
• Protein Structure Prediction: Provides residue depth constraints that can enhance protein structure prediction from sequence.
• Homology Modeling: Supplies residue positioning information to improve homology models.

Methodology:

Support vector regression (SVR) trained with eight sequence encoding schemes, with performance evaluated by 5-fold cross-validation.