The (Q)SAR methods and the in silico methods are often used to “predict” chemical-physical and toxicological properties of molecules with a known structure. The chemical industry is constantly evolving, and it needs fast and reliable methods to study the new molecules.

Just think of the 15,000 molecules registered every day (5.5 million per year!) on the Chemical Abstract Service (CAS), which are added to the 144 million already registered in the past. Of these molecules, around 95% is completely (or almost completely) lacking of toxicological data.

The need to develop fast and reliable methods to fill the information gap is therefore clear. In this context, The king of in silico methods is the (Q)SAR (Quantitative Structure-Activity Relationship) which, like Read-Across, is based on a general chemical concept coined by Carvin Hansch: “the chemical structure of a molecule influences its chemical-physical properties and its biological activity and, consequently, similar compounds have similar behavior”.

The (Q)SAR methods are applied by applying particular mathematical models developed from a set of molecules with both chemical structures and activity/toxicity/ properties known. Starting from these compounds, by using complex algorithms, the purpose is to identify a mathematical function that correlates the chemical structure to the activity/property. At this point it is easy to understand that the identified function can be used to predict a certain property of molecules with a known structure but unknown activity.

In the regulatory field, the (Q)SAR methods are accepted, and often suggested, by various international authorities. In particular, in 2006, the European Commission issued the REACH regulation, where the use of alternative methods is suggested in order to minimize the use of animals for toxicity tests.

The applications of these methods are various; for convenience, the molecules on which these instruments are usually applied are divided into three broad categories:

1. Molecules that benefit human health: all types of drugs and some food ingredients are included. The (Q) SAR models are often used to optimize power, receptor specificity, pharmacokinetic profile and reduce the toxicity of these molecules.
2. Chemicals potentially dangerous for the environment: all the molecules that come into contact with the environment are potentially dangerous for the ecosystem. For this reason, (Q) SAR methods find application in the identification of dangerous molecules associated with the environment.
3. Substances used in chemical processes: in this context, (Q) SAR is used to identify certain properties of molecules used in chemical-industrial processes in order to maximize the efficiency of an operation. Classic examples are the prediction of features such as the critical micellar concentration of surfactants, thermal degradation data and data on metal oxidation by chemicals.

One of the biggest problems related to the world of (Q)SAR, but generally to all in silico techniques, is that a result is always obtained, but it is up to the end user to understand whether or not the obtained output is reliable.

First of all, the (Q) SAR models used must be scientifically valid, and to be so they must have precise peculiarities. These characteristics (5 in total, which we will not deepen) were defined by the scientific community in the early 2000s and officially published in 2007 by the Organization for Economic Cooperation and Development (OECD).

The user must then go deeper into the result obtained, always taking into consideration the limits of an approach that is in all respects of a statistical nature and, by its nature, may be wrong.

In fact, these methods should not be used as an absolute alternative to experimental data but must be inserted within a context of “weight of evidence” in order to support an evaluation process, for example a toxicological one.

It is therefore very important that these analyzes are performed and interpreted by experienced professionals!