Artificial neural networks have revolutionized the drug discovery process. Nevertheless, two handicaps associated with this class of machine learning methods still remain: a) their lack of interpretability and b) the inability to simultaneously include several stages of the drug discovery process. This chapter intends to demonstrate that computational models based on artificial neural networks can be used in a chemistry-friendly manner to accelerate the design of virtually new, potent, and safe therapeutics at the preclinical level. We report the first multi-scale model for quantitative structure-biological effect relationships based on an ensemble of artificial neural networks (ms-QSBER-EL). The purpose of this model was to simultaneously predict the antimalarial activity, cytotoxicity, and the pharmacokinetic properties of the chemicals. The model displayed accuracy higher than 90% in both training and test sets. The different molecular descriptors present in the ms-QSBER-EL model were interpreted from a physicochemical and structural point of view. Such interpretations permitted the extraction and selection of different molecular fragments that were assembled, leading to the design of ten molecules. Six of these molecules were predicted by the ms-QSBER-EL model as potent and safe antimalarial agents. The designed molecules complied with Lipinski's rule of five and its variants.