Advances in information technology have enabled the development of computerized (in silico) methods for biomedical research and the testing of chemical and biological substances. To date, most of these advances have been made in the field of toxicology. Numerous computational methods have been developed to predict the toxicity and physico-chemical properties of products (particularly for products subject to REACH regulation), and can thus play a role in replacing and/or reducing animal testing.

Examples of in silico methods used in toxicology

SAR or QSAR models

Structure-activity relationship (SAR) or quantitative structure-activity relationship (QSAR) modeling is based on available data on existing chemicals. These relationships are integrated into databases, using mathematical models and reasoning by analogy, by families of substances, to predict the toxic risk of a substance; these methods have been used, for example, to study skin sensitization, eye irritation, etc.

Read-across

The read-across method is used to predict the toxic properties of a substance whose chemical structure is very close to that of another substance whose properties are already known. It’s a fast and inexpensive method. The difficulty lies in assessing the characteristics likely to attest to the proximity of the two substances under consideration; however, the relevance of this assessment is enhanced by the quality of the databases. This technique is used in many fields: carcinogenicity, hepatotoxicity, aquatic toxicity, reproductive toxicity, skin sensitization, eye irritation, environmental toxicity.

PBPK models

Physiologically-based pharmacokinetic (PBPK) models are mathematical representations of the absorption, distribution, metabolism and elimination of chemicals in humans or other animal species. They are used for a variety of purposes, such as simulating product concentrations in the body. They can also help predict variations in sensitivity between individuals and at different stages of development, which cannot be adequately addressed by conventional animal tests.

To build and run these models, machine learning tools and databases are available, notably within the REACH framework, such as the RASAR database. In some applications, these tools and models have been shown to perform better than animal testing in predicting product safety.

Integrated systems

Other approaches seek to integrate information from several sources. Adverse Outcome Pathways (AOP), a method promoted in particular by the OECD, seeks to model the sequence of biochemical events in the event of exposure to a substance, taking into account the complex interactions of factors leading to toxicity.

Finally, Integrated Approaches for Testing and Assessment (IATA) enable decisions to be taken on the toxicity of substances using sources of information from different levels of approach: physico-chemical properties, QSAR and cross-reference methods, in vitro and in vivo toxicity tests, AOP.

Other “in silico” methods

Computer modeling of physiological function and disease

This is another field in which in silico methods can replace the use of animals, as knowledge accumulates. Once all the parameters of a physiological or pathological phenomenon have been described, it becomes possible to proceed with computer modeling, linking processes and enabling simulations that are not feasible in vivo, for example to determine the impact of modifying a given parameter on the evolution of the modeled system.

Artificial intelligence tools are increasingly mobilized for this purpose. For example, the AlphaFold system, developed by DeepMind and acquired by Google in 2014, can predict the three-dimensional structure of a protein from the linear sequence of its constituent amino acids, in a very short time (around two hours for a chain of a few hundred amino acids). AlphaFold should considerably speed up the determination of protein structures. This can help to formulate hypotheses on the biological functions of molecules, which can be tested by introducing mutations, or to search for small molecules capable of attaching themselves to one of the folded parts of the protein in order to block or disrupt its function, with a view to treating certain diseases or neutralizing pathogenic agents (click here).

Another example: in toxicology, artificial intelligence is mobilized to extract relevant information from the automatic analysis of all known data; an example has been developed in France on bisphenol S, identifying a risk of obesity (article here). The tool itself is not designed to prove the toxicity or harmlessness of a substance, but to point out effects that need to be studied. Without eliminating them completely, it could make it possible to “reduce animal testing by directing towards more targeted tests”.

Systems biology (or systems biology, or integrative biology)

It offers new analytical possibilities in a wide range of fields, based on mathematical and statistical analyses. It approaches organisms and systems in all their complexity, seeking to integrate the various interacting parameters (intracellular organelles, cells, gene and protein networks, cell-to-cell communications, etc.), on several levels of analysis (multi-scale integration), in order to produce an operating model of the entire system, and to understand the dynamic interactions between components of a living system (and between living systems interacting with the environment). It is then possible to simulate the effect of a drug, for example, by analyzing all the levels of organization of the system, modifying the desired parameters, and making predictions. Systems biology is intrinsically interdisciplinary.

Virtual reality and surgical simulation

These are just some of the applications of computer modeling that have been implemented at Strasbourg’s university hospital, the Institut hospitalo-universitaire: to find out more, click here.

The ” Visible Patient ” solution offers organ simulations for the majority of digestive, thoracic, urological and pediatric surgeries: to find out more, click here.

Another example is Microsoft’s Hololens system, an autonomous holographic computer that conjures up 3D objects in the physical world, with which we can interact, and which has a wide range of applications in the healthcare field: diagnostic aids, assistance with surgical operations, better training for doctors, etc.

For more information click here.