Session: Undergraduate Research Posters: Computational Chemistry:

Knowing a protein’s structure has always been an interest among scientists due to the important relationship it has with its function. Multiple computational methods, that include prediction and assignment algorithms, have been developed to assist the extensive and difficult experimental elucidation of protein structures. These algorithms are mainly based on chemical features and statistical analyses. With the purpose of creating a secondary structure assignment tool that is based on geometrical features, we studied a core protein database composed of 186 proteins with known structures. The protein’s backbone is considered a polygonal chain which can be described by geometrical concepts. Using Wolfram Language, curvature and torsion values were calculated for all the residues of each protein of the database employing their crystallographic coordinates. The compiled results showed a clear statistical relationship between curvature-torsion pairs of the residues and the secondary structure element they belong to. Based on this relationship, a scoring system was created to design a secondary structure assignment tool with curvature and torsion values as sole inputs. This algorithm showed a Q3 accuracy (helix, sheet, and coil) of 74.2% for the training database. Considering the limited size of its training database, this assignment tool also performed remarkably well with a testing database. The novelty and simplicity of this geometrical approach lie in the fact that curvature and torsion values implicitly contain chemical and structural information such as hydrogen bonds and the lateral chain properties. This method could be improved in future research by increasing the size of the database and refining the scoring system, or even including additional geometrical descriptors.

Laura Gil Pineda

Appearances