Escherichia coli remains as one of the principal pathogens that cause nosocomial infections, being these, medical conditions which are increasingly common in healthcare facilities. Escherichia coli is intrinsically resistant to many antibiotics, and multidrug resistant (MDR) strains have emerged recently. Chemoinformatics has been a great ally of experimental methodologies such as high throughput screening (HTS), playing an important role for the discovery effective antibacterial agents. However, there is no approach able to design safer anti-Escherichia coli agents, due to the multi-factorial nature and complexity of bacterial diseases and the lack of desirable ADMET (absorption, distribution, metabolism, elimination, and toxicity) profiles as a major cause of disapproval of drugs. In this work, we introduce the first multitasking model based on quantitative-structure biological effect relationships (mtk-QSBER) for simultaneous virtual prediction of anti-Escherichia coli activities and ADMET properties of drug/chemicals under many experimental conditions. The mtk-QSBER model was developed from a large and heterogeneous dataset of more than 37800 cases, exhibiting overall accuracies higher than 95% in both, training and prediction (validation) sets. The utility of our mtk-QSBER model was demonstrated by performing virtual prediction of properties for the investigational drug avarofloxacin (AVX) under 260 different experimental conditions. Results converged with the experimental evidences, confirming the remarkable anti-Escherichia coli activities and safety of AVX. Predictions also showed that our mtk-QSBER model can be a promising computational tool for virtual screening of desirable anti-Escherichia coli agents, and the present chemoinformatic approach could be extended to the search for safer drugs with defined pharmacological activities.