In collaboration with Herakles, a research program was set up to have, in the future, optimized tools for predicting the toxicity of High Energy Molecules (HEM). Because of the number of mutagenicity tests commonly use in the society, a variety of internationally agreed testing protocols on chemicals was set up, described by REACh (Registration, Evaluation, Authorization and Restriction of Chemicals) and OECD (Organization for the Economic Cooperation and Development). The first test described for the mutagenicity is the Ames test. Several methods can be used to predict toxicity such as (Quantitative) Structure Activity Relationship [(Q)SAR] and On-The-Fly QSAR (OTF-QSAR). The aim of this paper is to evaluate the reliability of the SAR and the OTF-QSAR predictions in accordance with various similarities used. The goal is to obtain a good prediction tool for the Ames test. The prediction tool is compared with softwares often used in the scientific community. It was found that the OTF-QSAR predictions were more accurate and increased the efficiency of predictions. The precision of the system increases as the degree of similarity increased. A new zone was discovered and the best similarity found and refined. At the end of the study, we validate the similarity information and the OTF-QSAR on HEM molecules.