Association of combined features of amino acid and protein withcrystallization propensity of proteins from Cytophaga Hutchinsonii

Abstract

Various features of amino acids have been so far associated with crystallization propensity of proteins. The majority of features generally represents a certain aspect of individual amino acid, for example, molecular weight of an amino acid. Meanwhile, a small portion of features, which represents a certain aspect of a whole protein, is also associated with crystallization propensity of proteins, such as protein length. Clearly, the features of individual amino acids distinguish themselves from the features of a whole protein. Therefore it would be more rationale to use the features, which combine both features of individual amino acids and a whole protein, to associate crystallization propensity of proteins, because the features of individual amino acids are not subject to amino acids‘ positions in a protein, for instance. In this study, each of three combined features of individual amino acids and a whole protein are associated with crystallization propensity of proteins from Cytophaga Hutchinsonii through logistic regression and neural network, and each of 535 features of individual amino acids is also associated with crystallization propensity of the proteins to serve as benchmark. The results show that the combined features have a good relationship with crystallization propensity of proteins from Cytophaga Hutchinsonii. This study provides the information that the combined features can be used for predicting crystallization propensity of protein.