Quantitative relationships between structure and physicochemical properties of natural amino acids using topological and quantum-chemical molecular descriptors
Sashikanta Sahoo, Minati Kuanar, Suhani J. Patel, B. K. Mishra · 2014
Nineteen physicochemical properties of natural amino acids are correlated by combination of a large number of molecular descriptors considering simple connectivity of atoms, their topological orientations in two dimensions and quantum-chemical molecular descriptors describing interactions among them. All the molecular descriptors are amalgamated to generate the principal components, which are developed from the contribution of each parameter to different extents. To obtain a significant correlation and, thereby an optimized regression model for prediction of the physicochemical parameters, successive exclusion of variable technique has been applied successfully. The physicochemical parameters, not reported for any amino acid earlier, are predicted by using these models. In biological systems, molecular recognition usually depends on the structural complementarities of different compounds or functional groups. The triplet nucleotide code of amino acid (AA) also stems on this molecular recognition. To get a quantitative approach to this aspect, conversion of the structural formula into numerical values, thus encoding structural information, has a major role. Attempts are being made to generate quantitative structure-activity relationships among the biological molecules with their physical, chemical or biological properties. Despite various difficulties in QSAR analysis, several successful results have been reported in the quantitative description of molecular structure. The first attempt in numerical coding of amino acids (AAs) was made by Sneath, 1 who assembled a matrix consisting mainly of qualitative variables related to the presence or absence of functional groups or various other features of AAs. By using multivariate analysis, he derived various parameters such as similarity index, dissimilarity index etc. A similar approach was made by Simon, 2 who derived ten indicators for AAs. These indices did not get much appreciation due to their inability to predict