Protein structure assessment using knowledge-based statistical potential function based on Ramachandran plots

Background: Protein structural data is of prime importance for researchers in various biological fields, andthe accuracy of the data can impact the significance of analysis results. However, native protein structuresare not always available, so predicted protein models are widely used in many cases. Generally, in predictingprotein structures, knowledge-based scoring methods based on statistical potential are used to select the moststable and native-like structure from the predicted models. For instance, the Modeller simulation tool uses"dope-score" to rank protein structures. According to a study by Anfinsen et al., the native structures havethe least amount of free energy in a set of simulated models.Materials & Methods: Methods based on statistical potential do not have the certainty and accuracy forcalculating molecular mechanics force field; in contrast, statistical potential functions are computationallypossible. For example, out of ۵۵ decoy sets created by QUARK and I-TASSER algorithms for CASP۱۱targets, dope-score allocates only ۱۵ minimum energies to native structures. Therefore, the need for researchto achieve better evaluation criteria remains essential. This project intends to design a new knowledge-basedscoring function based on PDB database information. Ramachandran diagrams are created for amino acidwindows with ۲ to ۷ residues. The probability of any angle occurring in each window using statisticalpotential and the Boltzmann function is converted into energy. The protein with the least energy is consideredas the best model.Results & Conclusion: Our method could successfully rank more native protein structures as the top ۵%among all modeled proteins in CASP۱۱ than dope-score function.

کلیدواژه ها:

Validation of structure ، protein structure ، Ramachandran ، statistical potential ، protein quality