Category | Representative Methods | Description |
Deterministic | Gradient Descent, Conjugate Gradient, L-BFGS | Used for local refinement after a good starting conformation is found |
Stochastic | Monte Carlo (Metropolis), Simulated Annealing | Random sampling to escape local minima |
Evolutionary | Genetic Algorithms, Particle Swarm Optimization | Population-based global search |
Molecular Dynamics (MD) | Energy minimization followed by time-evolution | Uses Newtonian mechanics to sample conformational space |
Machine Learning / AI | AlphaFold, RoseTTAFold | Uses neural networks trained on structural databases to predict folding without explicit energy minimization |
Parameter | Description | Impact |
Hydrophobicity pattern | Degree to which residues avoid water | Drives hydrophobic collapse — core formation in globular proteins |
Amino acid composition | Ratio of polar, nonpolar, charged, aromatic residues | Determines secondary structure preference and stability |
Residue order / motifs | α-helix or β-sheet forming propensities (e.g., Ala, Leu favor helices; Val, Ile favor β-sheets) | Controls local folding patterns |
Chain length (N) | Number of residues | Affects folding rate and number of possible conformations (exponentially) |
Disulfide bonds (Cys–Cys) | Covalent cross-links | Strongly stabilize tertiary structure and folding cooperativity |
Proline and glycine content | Structural disruptors or flex points | Influence turns and loop flexibility |