Frankfurt After the IT industry, the pharmaceutical industry is now also increasing its investments in artificial intelligence (AI). This is shown by a number of takeovers and alliances completed in the past few weeks. The pharmaceutical companies expect a more targeted and faster drug search.
So far, computer- and data-supported procedures have not yet met the great expectations in medicine, because the biological processes in the body are very complex. However, pharmaceutical managers will soon see the technology on a new level. “We are reaching a stage where progress in the field of AI must be combined with our expertise in drug development,” says Biontech boss Ugur Sahin.
The successful German biotech company recently pushed ahead with the acquisition of the British AI start-up Instadeep, with the help of which novel vaccines and biopharmaceuticals against cancer and infectious diseases are to be found.
Almost at the same time, Bayer announced the acquisition of the Scottish company Blackford Analysis, which deals with the use of artificial intelligence in radiology. The Leverkusen group also agreed a partnership with the Google parent company Alphabet. The aim is to accelerate early drug discovery using machine learning models in the Google cloud.
Top jobs of the day
Find the best jobs now and
be notified by email.
In November, Sanofi agreed to its third potentially multi-billion dollar alliance in the field in less than a year. In this case, the French pharmaceutical company allied itself with the US-Chinese company Insilicio, which specializes in AI-based drug discovery. Previously, Sanofi had already sealed larger alliances with the British Exscientia and the US company Atomwise.
Pharmaceutical managers rate artificial intelligence as the most important disruptive technology for the industry, according to a new survey by the British analysis and consulting firm Global Data. Several hundred young start-ups now offer AI systems and processes, and some of these companies have even entered the pharmaceutical business themselves.
Big data analyzes and machine learning methods promise advantages on several levels for drug development. For example, in the search for molecular targets and in the selection of potential drug substances. The huge amounts of data from genetic analysis can no longer be managed without computer-aided processes.
The technology is also becoming increasingly important for the analysis of proteins. The Alphabet subsidiary Deepmind caused a stir last year with a system that can calculate the spatial structure of proteins based on the underlying gene sequences. The technology is therefore suitable for significantly accelerating basic biochemical research. With their help, complex proteins can be designed, but also smaller, chemically synthesizable substances.
Despite all the advances and excitement, it remains difficult to gauge the power of technology in drug discovery. In the past, the field of work has proven to be notoriously prone to hype and exaggerated expectations.
As early as 1981, the US magazine “Forbes” celebrated computer-based drug development as “the next industrial revolution”. Two decades later, young genome analysis companies promised a radical change in drug development based on their genetic data. Both turned out to be illusions.
Pharmaceutical industry is investing billions of dollars in AI
Since the middle of the last decade, the industry has already invested more than $1 billion in direct payments in alliances with AI specialists and committed another approximately $40 billion in performance-based bonuses. But even that did not bring any clearly recognizable improvement in research efficiency.
Although the global pharmaceutical industry’s output of new active ingredients grew in the first half of the last decade, it has stagnated at the increased level since then and even fell by around a quarter last year.
A number of companies that have already loudly pounced on digitization strategies in the last decade have actually lost ground in competition since then. Novartis boss Vas Narasimhan, for example, who already thought the Basel group was on the way to becoming a “medical and data science company” in 2018, is currently struggling with weaknesses in the product range and has to reorganize the pharmaceutical business.
At the beginning of 2017, the Darmstadt-based Merck group sealed a big data alliance with the US data specialist Palantir in order to develop, market and provide medicines more quickly. However, the range of product candidates in Merck’s pharmaceutical pipeline has not increased since then, but has been halved.
>> Read about this: Merck is streamlining pharmaceutical research – jobs in the USA are being eliminated
Investors are skeptical about the sector despite the recent spate of deals. Listed pharma AI specialists such as Relay Therapeutics, Exscientia or Benevolent AI have lost between 30 and 60 percent in value since the end of 2021.
The complexity of biology
A central challenge for AI systems in pharmaceutical research continues to arise from the enormous complexity of the biological processes in the human body. So far, they have not been able to be reliably mapped in computer models and algorithms because there is often a lack of data and knowledge of the clinical relevance of data.
The bioinformaticians Andreas Bender and Isidro Cortés-Ciriano see this as an Achilles heel when using AI systems in pharmaceutical research. “To really advance the field, we need a better understanding of biology,” they write in a comprehensive analysis of AI-based research approaches.
A typical problem arises from the fact that computer models are now very good at calculating whether drug candidates can dock onto a specific target molecule. However, this says little about its suitability as a drug.
Rather, numerous other factors are also important, such as whether they survive in the metabolism at all, in which cells they are absorbed, how they interact with other molecules and processes in the cells, how they are broken down or enriched.
According to some estimates, there are now well over 100 AI-based products in clinical tests. But these are almost always very early studies that have not yet been able to provide any evidence of basic effectiveness.
But it is foreseeable that both the database and the performance of the AI processes will gradually improve. That’s why the big pharmaceutical companies are now staking out their claims in the area.
More: After the Corona boom, the pharmaceutical industry is facing a dip in growth.