From molecular design to drug repositioning, in silico research uses artificial intelligence and computational models to reduce time, cost and failure in the development of new therapies.
Keep this term in mind: in silico. Because that’s what you’re likely to hear when the cure for cancer is discovered, or the antibiotic that solves resistant infections is found, or the cure for Alzheimer’s. We could say that this strange Latinism was coined by exclusion, to give a name to pharmaceutical research that is not in vitro (i.e. in a test tube) and neither alive (therefore on living organisms) but proceeds thanks toArtificial intelligencein the silicon of chips and in the computational models of computers.
The drug laboratory of the future, in fact, has no white coats or test tubes. Instead of the controlled chaos of the laboratory bench, there are codes and algorithms: we research the medicines we will use in the coming years, but we don’t mix chemistry. The Capgemini Research Institute’s recent report on the biopharmaceutical industry predicted that in a decade the 60 percent of new drugs will have been designed byAI. Far from science fiction, AI research is already starting to change the history of medicine.
De novo drug design and drug repurposing
«Drug design in silico it is not only useful for creating molecules from scratch, but is also useful for finding new uses for existing ones”, Antonio Lavecchia, full professor of pharmaceutical chemistry at the University of Naples Federico II and author of the book, explains to Panorama Applied Artificial Intelligence for Drug Discovery. «On the one hand, thanks to computers and AI, researchers design new molecules, simulating how they could bind to a protein involved in a disease: this is the so-called de novo drug designin which the drug is created “on the computer” and only afterwards, if promising, is it produced and tested in the laboratory.”
But there is an equally important approach: the drug repurposingthe repositioning of drugs. «In this case» continues Lavecchia «we are not starting from scratch, because the AI and the models in silico they analyze large quantities of data on already approved medicines, on their chemical structure and mechanisms of action, to understand if they can be effective on diseases other than the original ones”. A concrete example was that of baricitinibdeveloped for rheumatoid arthritis and then repurposed for the treatment of Covid-19.
Antibiotic resistance and new molecules
«In the field ofantibiotic resistanceparticularly crucial today, there are already concrete examples of the impact of artificial intelligence on research”, underlines Eugenio Santoro, head of the Unit for research in digital health and digital therapies of the Irccs Mario Negri Institute. «I am referring, for example, to studies on halicin And abaucinaexperimental antibiotics identified – in the first case – through repositioning strategies and in the second by new analyzes carried out on millions of compounds”.
Halicin, named after Hal, the computer of 2001: A Space Odysseyoriginally developed for diabetes, has proven capable of defeating many bacteria, including the dreaded one Acinetobacter baumanniiresistant to all antibiotics known so far. He’s not the only one: the Clostridium difficilefor example, causes an infection that is often lethal in the elderly and which generates hospital costs of around 5 billion a year in Europe.
From discovery to clinical trials
«Boston biotech Insilico Medicine also used AI platforms to generate a drug candidate for idiopathic pulmonary fibrosis», explains Lavecchia. «This compound, called rentosertibwent from discovery to early stage clinical trials in approximately 30 months, a much shorter time than traditional development cycles.” In 2024, Insilico completed patient enrollment for the Phase 2 study in China and published on Nature Biotechnology detailed results. “Preliminary data indicate that rentosertib was well tolerated and shows initial signs of efficacy.”
Accelerations and savings more necessary than ever: today the 90 percent of drugs that reach the last phase of testing fail to achieve their objective and each new molecule that arrives on the market, after 12-15 years of testing, costs on average 2.8 billion dollars. A study published in Drug Discovery Today However, in June 2024 it showed that molecules discovered with AI have an initial success rate of80-90 percent.
When AI arrives at the patient’s bedside
Some drugs “repositioned” thanks to AI have already reached patients. This is the case of clonazepamused for an American girl suffering from the very rare DeSanto-Shinawi syndrome. Through the platform BabyFORceAI selected the most promising one among existing medicines to increase the production of the gene involved. The results are still preliminary, but the little girl has started to pronounce her first words and crawl.
Big tech, regulators and the future of medicine
Even the technological giants have entered the field. An AI system called C2S-Scale 27Bdeveloped by Google DeepMind together with Yale University, generated a new hypothesis on the cellular behavior of cancer, then validated experimentally. “With further preclinical and clinical testing, this discovery could reveal a promising new avenue for developing therapies,” Sundar Pichai wrote.
Giuseppe Remuzzi is convinced of the importance of AI also in diagnoses, citing a study by New England Journal of Medicine: AI reaches the85.5 percent correct diagnoses versus 20 for clinicians. Regulatory agencies, from the EMA to the FDA, have already published common guidelines for the responsible use of AI in medicines. Research is also underway in Italy: university groups are already synthesizing new molecules against antibiotic resistance.
Maybe test tubes will not disappear completely, but they will no longer be the place where intuition is born. That will take shape first, in the chips. Because, as Hippocrates wrote, “life is short, art is long”. And today the art of medicine has learned to think in silico.
