Disease vector insects rely on chemosensors to locate hosts, find mates and choose where to lay their eggs. Currently, the most efficient method of preventing and controlling the outbreak of insect-borne diseases is the use of Insect Repellents (IRs). However, current IRs have significant drawbacks and do not meet the necessary conditions, such as protecting a broad spectrum of mosquitoes; many have unpleasant odors or produce unpleasant sensations on the skin. Some of them are even carcinogens. Therefore, the need for new, more effective, safer, and longer-lasting broad-spectrum IRs than conventional IRs is evident. Here, classifiers for predicting IRs will be developed using QuBiLS-MIDAS 3D-molecular descriptors and shallow machine learning techniques. We intend to introduce, the ability of QSAR (Quantitative Structure-Activity Relationships) models to describe the interaction of IRs with the olfactory response of the six sensilla of the mosquito Culex quinquefasciatus as well as with repellent activities by using four datasets that take into consideration the two most relevant IR scaffolds: carboxamides and plant-derived compounds with repellent effect on A. aegypti and also A. gambiae and the two most common species of cockroach (Blattella germanica and Periplaneta americana). A non-commercial and cross-platform software, named “SiliS-PAPACS” is developed for the IRs-prediction (http://tomocomd.com/apps), in which all the best models developed are implemented. This software is used to screen datasets containing diverse chemotypes like chemicals and/or FDA-approved drugs, like Malaria box library. From that virtual screening, we report 28 novel virtual leads (new IR chemical scaffold) that may have potential IR activity. The results suggest that the proposed method will be an excellent computer-assisted system that could increase the chance of finding new insect control agents and assist those researchers in screening and/or designing new chemotype IRs.