During the last years, the 3Rs (Replacement, Reduction, Refinement) principle is increasingly taken into consideration in setting up integrated testing strategies. As such, in silico methods received substantial attention, which stimulated their development and made them become more interesting for the assessment of chemical hazards. In silico tools are essentially computer models, able to make predictions for a non-evaluated compound based on knowledge extracted from a collection of structurally related substances with experimental toxicity data. Progressively acknowledged by regulatory bodies, in silico tools are gaining importance in toxicology not only as a first tier screening tool, but also for complementing in vivo and in vitro test results (for example, Buist et al. 2013; Nendza et al. 2013; Schilter et al. 2014; Scholz et al. 2013). Their widespread use, however, remains limited due to (1) the non-flexibility of the current regulatory framework, strictly describing the required experimental tests, (2) the oversupply of computer models while often uncertainty exists as to which model (combination) is most suitable to assess a given (type of) substance for a particular endpoint, and (3) the rather poor predictive capacity for toxicological endpoints other than Ames mutagenicity (Barber and Myatt 2016).

Therefore, the most promising application of in silico tools today remains its use in priority setting upon screening of a large number of compounds. The general public is exposed, intentionally or not, to a large variety of different substances, sometimes not or not recently evaluated for their safety. Environmental pollutants or food contaminants are evident examples of non-intentional exposure to predominantly non-evaluated substances. A detailed characterization of the complete toxicological profile of all these substances is not feasible from an economic and ethical (animal welfare) point of view. In silico tools, however, can provide substantial help in assigning priority to those substances for which a comprehensive safety evaluation is most urgently needed.

In a recent study, we illustrated the potential of in silico tools for such priority setting in the field of printed paper and board food contact materials (FCM) (Van Bossuyt et al. 2017). Paper and board FCM, like other non-plastic FCMs, can contain a large number of non-evaluated substances (European Parliament 2016; Liu et al. 2016; Muncke et al. 2014; Van Bossuyt et al. 2016). Several food crises have confirmed that FCM substances can migrate into food and drinks, subsequently causing unwanted exposure of the consumer to potentially harmful substances (European Commission 2016; EFSA 2011). As a result, health concerns have been raised which are justified especially since migration from FCM is estimated to be the main source of food contamination, quantitatively exceeding most others—including pesticide residues—by a factor of 100–1000 (Grob et al. 2006).

The study focused on Ames mutagenicity, an important toxicological endpoint related to serious adverse health effects including carcinogenicity. For Ames mutagenicity, a number of valid in silico models are available to make substantiated predictions. Four different in silico tools were used to select currently non-evaluated printed paper and board FCM substances that most likely exhibit mutagenic properties. In particular, the substances identified as being of ‘highest priority’ need immediate further investigation. By identifying substances of highest concern, the resources available for experimental testing can be attributed in a more efficient way. Indeed, it would be impossible to carry out elaborate toxicological investigations for hundreds of chemicals in the course of an acceptable time span.

Similarly, in silico tools can be of particular interest to screen ‘non-intentionally added substances’ (NIAS) migrating from FCM. These NIAS include impurities, oligomers and degradation products (Muncke 2011; Nerin et al. 2016). NIAS typically represent the larger part of the migrants and their exact composition is often unknown (Grob 2014). In this context, we recently reported on the use of in silico tools as part of a strategy to identify non-authorized chemicals of genotoxic concern found to migrate from plastic baby bottles used as alternative to bisphenol A-containing polycarbonate baby bottles (Mertens et al. 2016). In addition, in this case study, the importance of in silico tools for prioritization of FCM substances was clearly demonstrated.

This prioritization strategy based on in silico methodology can also be applied in several other domains where there is a need to identify priority substances requiring (geno)toxicological evaluation. Actual examples include compounds used in tattoo inks, permanent make-up, printed baby napkins and sanitary towels, medical devices, textile products and nanomaterials. In all these cases, human health safeguarding can be realized already to some extent without the use of experimental (animal) systems.