From DNA, the genetic code is translated into proteins being these macro molecular structures in charge to execute specific cellular functions. All proteins are composed by the combination of the same 20 amino acids, indeed, it is the 3D shape and detailed structure that differentiate a protein from another and so their functions. Determining protein structure is fundamental to understand their biological function; modify and alter their structures is the way in which drugs can block or activate a protein activity and doing so preventing or treating diseases. In our lab we are understanding protein and DNA structures to the level of atomic resolution contributing to the biomedical research. We process with python the information of hundred thousands 3D coordinate structures already known and predict patterns that can be used to determine structure of unknown proteins. To achieve this objective, proteins are first reduced to a series of specific 3D vectors (Characteristic Vectors [1]) and then the geometrical relationships among them are extracted. We make use of biopython and python-igraph libraries to model each structure vectors in a network graph and we use supervised and unsupervised learning to classify and predict structure features. Python-igraph is a wrapper to a C code implementation of a set of graph algorithms of classical use in network analysis.

[1] Sammito M1, Millán C, Rodríguez DD, de Ilarduya IM, Meindl K, De Marino I, Petrillo G, Buey RM, de Pereda JM, Zeth K, Sheldrick GM, Usón I. Exploiting tertiary structure through local folds for crystallographic phasing. Nat Methods. 2013 Nov;10(11):1099-101. doi: 10.1038/nmeth.2644. Epub 2013 Sep 15.