In his report on artificial intelligence, submitted in March 2018 (see Tangente 181, 2018), Cédric Villani identifies healthcare as a strategic sector. This is not the future but the present: more and more initiatives using statistical models and artificial intelligence are emerging.
In silico trials ----------------------
The safety and efficacy of a drug are currently established using in vitro models in test tubes, in vivo animal models and, above all, clinical trials in humans. In silico modelling and simulation are emerging as a new means of assessment (see Mathématiques et Médecine, Bibliothèque Tangente 58, 2016). The expression "in silico" derives from the word "silicon", a basic component of computers, and denotes purely computer-based studies.
How does it work? A model is created using artificial intelligence, combining mathematical models of biological mechanisms in real patients with computer simulations of virtual patients, so as to incorporate all the available information about these mechanisms. In practical terms (if that is the right phrase), "virtual patients" are "constructed" by programming all the biological phenomena known from previous studies, with varying degrees of sophistication—from ten to several hundred equations, using differential equations and advanced regression models. The disease of interest is then "artificially created" by adjusting the parameters, again on the basis of what is known about the condition. The effect of the treatment is simulated in the same way, drawing on initial observations in animals or humans and on existing knowledge of cellular mechanisms, biological reactions and so on. We can then compare the outcome for a treated virtual patient with that for the same virtual patient without treatment and see whether the drug is effective.
The entire model—that is, all its equations—must be validated against the initial clinical data: the statistical model must produce the same results as those observed. This validation stage is essential. The aim of in silico trials is therefore to speed up the assessment of new drugs.