Pesticides represent extremely useful chemical and biochemical agents for the prevention of crop losses and disease in humans. They act by destroying, repelling or mitigating pests. Insects constitute one of the pests which are very difficult to control. With the pass of the time, insects have become resistant to pesticides because the current insecticides are designed to act through one mechanism of action. For this reason, there is an increasing need for the design of more potent and versatile insecticides. The present study is focused on the development of a fragment-based approach for the in silico discovery of multi-target insecticides from a heterogeneous database of compounds. The present methodology was based on a QSAR discriminant model which classified correctly more than 90% of insecticides and inactive compounds in both, training and prediction series. Also, it permitted the automatic and efficient extraction of fragments responsible of insecticidal activity against several mechanisms of action and new molecular entities were suggested as possible multi-target insecticides.