Protein tertiary structure prediction using hidden Markov model based on lattice.

Area of Science:

• Computational biology
• Structural bioinformatics
• Biophysics

Background:

• Protein structure prediction from amino acid sequences is a critical challenge in computational biology.
• Protein tertiary structure dictates biological function and is determined by the amino acid sequence through protein folding.
• Existing methods often rely solely on primary or secondary structure information.

Purpose of the Study:

• To develop a novel fold recognition method for protein tertiary structure prediction.
• To integrate hidden Markov models (HMMs) with 3D coordinates of amino acid residues.
• To explore the utility of Bravais cubic lattices in modeling amino acid residue paths.

Main Methods:

• A novel fold recognition method employing hidden Markov models (HMMs) and 3D coordinates.
• Introduction of states based on Bravais cubic lattice basis vectors (simple cubic, body-centered cubic (BCC), and face-centered cubic (FCC)).
• 10-fold cross-validation on a 42-fold SCOP dataset, comparing against sequence- or secondary structure-based HMM methods.

Main Results:

• The proposed method, particularly the face-centered cubic (FCC) lattice model, demonstrated superior accuracy compared to existing methods like SAM, 3-HMM optimized, and Markov chain optimized.
• The integration of extensive 3D spatial data significantly enhanced model performance.
• The FCC-based HMM approach outperformed methods relying solely on primary or secondary protein structures.

Conclusions:

• The novel hidden Markov model approach utilizing 3D coordinates and Bravais cubic lattices offers a significant advancement in protein fold recognition.
• The face-centered cubic lattice model shows particular promise for accurate tertiary structure prediction.
• Incorporating 3D structural information is crucial for improving the performance of protein structure prediction algorithms.