Integrated In Silico, In Vivo, and Deep Learning Approaches in the Discovery of Novel Candidate Molecules for Aedes aegypti Control.

The mosquito Aedes aegypti is a primary vector responsible for transmitting major arboviruses, including dengue, Zika, chikungunya, and yellow fever. Increasing resistance to conventional synthetic insecticides, combined with their well-known environmental drawbacks, underscores the urgent need for more selective, sustainable, and effective strategies for vector control. Chalcones have been previously identified by our research group as a promising chemical class of larvicidal agents, with preliminary evidence for distinct mechanisms of action. More recently, an additional strategy for integrated control of A. aegypti in its adult stage has emerged through the inhibition of blood feeding, particularly via agonism of neuropeptide Y-like receptor 7 (NPYLR7). In this context, this multi-pronged investigation was conceived as a stage-specific discovery framework addressing distinct biological vulnerabilities of A. aegypti. Specifically, the study aimed to: evaluate the larvicidal potential of chalcones through integrated in silico and in vivo approaches targeting juvenile hormone transport; apply deep learning-based high-throughput virtual screening (HTVS) as a candidate-prioritization strategy for identifying chemically plausible NPYLR7 agonists associated with blood-feeding inhibition; and finally generate novel NPYLR7-oriented molecular scaffolds using DeSAO ("de novo drugs using Simulated Annealing Optimization)" algorithm as a hypothesis-generating de novo design methodology. These strategies were intentionally pursued as complementary, rather than convergent, discovery axes reflecting the distinct biological requirements of larval and adult mosquito control. Initially, a classical docking-based virtual screening of 1070 chalcones from the PubChem database was conducted on the A. aegypti juvenile hormone-binding protein (mJHBP), a hemolymph-circulating protein involved in hormonal regulation of larval and adult development. Docking calculations revealed several analogues with favorable predicted binding energies. Three halogenated chalcones were then commercially acquired for experimental larvicidal assays, which identified 4'-chloro-4-methoxychalcone (2c) as the most active compound after 72 h exposure. In parallel, the Machine Learning driven HTVS and the DeSAO workflow independently identified and prioritized novel molecular scaffolds with predicted NPYLR7 agonist activity, generating chemically plausible candidates for subsequent experimental evaluation of blood-meal inhibition in adult mosquitoes. Collectively, the results indicate that halogenated chalcones with moderately sized substituents may serve as promising larvicidal candidates, while HTVS and DeSAO provide complementary, chemically diverse architectures for future evaluation in blood-meal control assays. Taken together, these findings reinforce the value of integrating computational, Machine Learning, and experimental methodologies within a unified pipeline, enabling both validated larvicidal discovery and biologically grounded candidate prioritization for adult mosquito control.