In the middle of the last decade, dozens of startups have rushed promising to revolutionize the field of drug discovery using artificial intelligence, by reducing the time required for drug development and reducing the cost, which reaches an average of about two billion dollars per new drug. These promises attracted major pharmaceutical companies such as Bristol-Myers Squibb and Sanofi, which signed multibillion-dollar deals, hoping to access more rapid and effective drugs.
But after more than ten years, experts are wondering: Where are the drugs promised by artificial intelligence? So far, no drug discovered via these technologies has received regulatory approval, and many of the first trials have failed, according to a report by the British Financial Times.
Some of the companies that promoted these promises faced financial crises due to declining investments in the biotechnology sector, including the British company Benefolentai, which aroused widespread enthusiasm when it was founded, but lost more than 99% of its market value before it was delisted from the stock exchange and merged into a Japanese company last March.
Alex Gavoronkov, CEO of the artificial intelligence drug discovery company Insilico, says: "to claim that you have the hen that lays golden eggs, you have to offer some really golden eggs. And if you don't, your value declines very quickly”.
Huge investments and renewed hopes
The idea of developing medicines with artificial intelligence has attracted investors because the pharmaceutical sector is slow and expensive, and is a promising area for change. According to the data, the financing of drug discovery companies with artificial intelligence increased from 30 million dollars in 2013 to 1.8 billion dollars in 2021.
The boom of generative artificial intelligence since the launch of ChatGPT in late 2022 has revived hope, and investors have returned to bet on new companies, more sophisticated technologies, innovative ways of collecting and analyzing biological data, as the first tools were not powerful enough, but future generations are capable of achieving.
However, the challenge remains, because we know little about the interactions of our cells, and we are unable to measure many of the basic vital processes, depriving models of sufficient data to achieve tangible achievements.
Veteran chemist Darren Green, who has worked for more than 30 years at the famous pharmaceutical company GSK, says: “Drug discovery is perhaps the most difficult thing that a person tries to do.. Whenever we get new tools, we find new obstacles, " he said, adding that failures are a frequent part of the industry.
Repeated failures and scientific challenges
For decades, the drug development sector has pinned its hopes on technologies such as structural biology in the Sixties, computational chemistry in the eighties, and the Human Genome Project at the turn of the millennium, but the results have come slower than expected. In the process of drug discovery, research begins with identifying a biological target, such as a genetic mutation or a hormonal receptor, and then searching for a molecule that can interact with it to treat the disease. The problem is that about 90% of the candidate drugs fail in clinical trials, while they succeed on paper in theory.
This is where the role of artificial intelligence stands out in accelerating the search within huge databases of molecules. But experts warn against overdoing it. Peter Coveney, professor of Chemistry at UCL University College London, said:” It is a mistake to think that any computer technology will suddenly be able to solve all the problems, for example, toxicity, i.e., side effects of drugs, is one of the most difficult things to predict.
Greater data and new opportunities
The latest transformations came with the launch of the alphafold2 model from Google DeepMind in 2023, as it was able to predict the shape of proteins with unprecedented accuracy, and these developments may reset the clock to zero, until a new cycle of promising attempts begins.
Today, companies such as Insitro, Recurrence, and Lila Sciences are trying to build “scientific factories” to produce large-scale biological data via automation and computer vision technologies.
“We need half a dozen major alphafold-sized developments to make a real leap in drug discovery,” says Max Gaderberg, head of artificial intelligence at Google's Isomorphic Labs.
The future is in the hands of big companies
Startups in the field of artificial intelligence drug development are struggling to prove the feasibility of their work, plans, and vision. Still, analysts believe that tech giants such as Google and Alphabet, thanks to their huge resources and supercomputing, may be the best able to lead this transformation.
Sanjeev Patel, CEO of Relay Therapeutics, sums up the matter: “This process can take many years without results, but only large companies can afford to wait until the desired achievement comes,” he said.
Decisive results are still lacking, but many agree that what is happening today may be the prelude to a pivotal stage in which artificial intelligence is transitioning from theoretical promises to practical tools that could change the established rules for discovering treatments and saving lives.