Development of a quantitative structure activity relations (QSAR) model to guide the design of fluorescent dyes for detecting amyloid fibrils

Abstract

Quantitative structure activity relationship (QSAR) studies were performed on a set of polymethine compounds to develop new fluorescent probes for detecting amyloid fibrils. Two different approaches were evaluated for developing a predictive model: part least squares (PLS) regression and an artificial neural network (ANN). A set of 60 relevant molecular descriptors were selected by performing principal component analysis on more than 1600 calculated molecular descriptors. Through QSAR analysis, two predictive models were developed. The final versions produced an average prediction accuracy of 72.5 and 84.2% for the linear PLS and the non-linear ANN procedures, respectively. A test of the ANN model was performed by using it to predict the activity, i.e., staining or non-staining of amyloid fibrils, using 320 compounds. The five candidates whose greatest activities were selected by the ANN model underwent confirmation of their predicted properties by empirical testing. The results indicated that the ANN model potentially is useful for facilitating prediction of activity of untested compounds as dyes for detecting amyloid fibrils.

Key words::

• amyloid fibrils
• artificial neural network
• cyanine dyes
• fluorescent probe design
• QSAR

Acknowledgment

This work was supported by Science and Technology Center in Ukraine, Project 5281 “Development of fluorescent dyes for detection of oligomeric amyloid intermediates in neurodegenerative diseases.”

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.