Indexing Natural Products for their Antifungal Activity by Filters-based Approach: Disclosure of Discriminative Properties

Abstract

Background: A considerable worldwide increase in the rate of invasive fungal infections and resistance toward antifungal drugs was witnessed during the past few decades. Therefore, the need for newer antifungal candidates is paramount. Nature has been the core source of therapeutics for thousands of years, and an impressive number of modern drugs including antifungals were derived from natural sources. In order to facilitate the recognition of potential candidates that can be derived from natural sources, an iterative stochastic elimination optimization technique to index natural products for their antifungal activity was utilized.

Methods: A set of 240 FDA-approved antifungal drugs, which represent the active domain, and a set of 2,892 natural products, which represent the inactive domain, were used to construct predictive models and to index natural products for their antifungal bioactivity. The area under the curve for the produced predictive model was 0.89. When applying it to a database that is composed of active/inactive chemicals, we succeeded to detect 42% of the actives (antifungal drugs) in the top one percent of the screened chemicals, compared with one-percent when using a random model.

Results and conclusion: Eight natural products, which were highly scored as likely antifungal drugs, are disclosed. Searching PubMed showed only one molecule (Flindersine) out of the eight that have been tested was reported as an antifungal. The other seven phytochemicals await evaluation for their antifungal bioactivity in a wet laboratory.

Keywords: Antifungal; bioactivity index; chemoinformatics; data base; ligand-based screening approach; natural products..