Multitarget inhibitors of insect chitinolytic enzymes are promising sources of green insecticides. Machine learning (ML) is an emerging virtual screening method that can accelerate drug discovery and reduce costs. Taking advantage of the data from our previous high-throughput screening work, we established a strategy integrating ML and molecular docking to screen a large natural product library (17 600 compounds) to identify novel multitarget inhibitors of four chitinolytic enzymes from the insect Ostrinia furnacalis (OfChtI, OfChtII, OfChi-h, and OfHex1). 3,5-Di-O-caffeoylquinic acid and γ-mangostin were identified as inhibitors of all of these enzymes with Ki values at the μM level. Moreover, they showed significant biological activities against lepidopteran pests. Transcriptomic analysis of compound-treated insects suggested the physiological relationship between these compounds and chitinolytic enzymes. This study highlights the potential of ML for insecticide discovery and provides novel and low-cost scaffolds of multitarget inhibitors against insect chitinolytic enzymes as potential pesticide leads.
Keywords: chitinase; inhibitor; insecticide; machine learning; natural product; virtual screening.