Novel amino acids indices based on quantum topological molecular similarity and their application to QSAR study of peptides

Abstract

A new source of amino acid (AA) indices based on quantum topological molecular similarity (QTMS) descriptors has been proposed for use in QSAR study of peptides. For each bond of the chemical structure of AA, eight electronic properties were calculated using the approaches of bond critical point and theory of atom in molecule. Thus, for each molecule a data matrix of QTMS descriptors (having information from both topology and electronic features) were calculated. Using four different criterion based on principal component analysis of the QTMS data matrices, four different sets of AA indices were generated. The indices were used as the input variables for QSAR study (employing genetic algorithm-partial least squares) of three peptides' data sets, namely, angiotensin-converting enzyme inhibitors, bactericidal peptides and the peptides binding to the HLA-A*0201 molecule. The obtained models had better prediction ability or a comparable one with respect to the previously reported models. In addition, by using the proposed indices and analysis of the variable important in projection, the active site of the peptides which plays a significant role in the biological activity of interest, was identified.