QSPR study of viscoplastic properties of peptide-based hydrogels.

In this study, the power of machine learning was harnessed to probe the link between molecular structures of peptide-based hydrogels and their viscoplastic properties. The selection of compounds was attempted in accordance with the prescribed full list of peptide-based materials exhibiting hydrogel functionality in the literature. In this pursuit, a complete set of molecular descriptors and fingerprints was considered - accounting for an entry of size 17,968 for each peptide-based structure analyzed. The elastic and viscous moduli response of materials were mapped over a wide frequency spectrum in the range [0.1-100] (rad/s). In general, the results indicate that the frequency-dependent mechanical response of peptide-based hydrogels is statistically correlated with its (inter)molecular attributes, such as charge, first ionization potential (or equivalently electronegativity), surface area, number of chemical substrates, bond type, and intermolecular interactions. The performance of several (supervised) soft computing techniques was measured, for our quantitative structure property relationships model. In addition, the hypothesis of mapping our databank to a new system of principal components was tested, by using an unsupervised methodology, which resulted in enhancement of the prediction accuracy. In terms of significance, the present article provides the first report of frequency-dependent elastic and viscous moduli, for a set of 70 peptide-based formulations with hydrogel functionality.Communicated by Ramaswamy H. Sarma.