Digital Health

New medicine thanks to AI?

In times when it is becoming increasingly difficult to discover new pharmaceuticals, the use of artificial intelligence (AI) also gives hope in this area. By analysing large amounts of data and developing predictive models, AI can significantly speed up the research and development process of new drugs. We shed light on why this could help cancer patients and what other opportunities, but also risks, arise.

KI bei der Arzneimittelherstellung

Anyone who discovers a mouldy packed sandwich in their child's school bag after six weeks of summer holidays is likely to be at least briefly annoyed by this accident. But one could also look at it differently: In science, similar "accidents" have already led to astonishing discoveries. The Scottish physician and bacteriologist Alexander Fleming for example found penicillin by accident. When he left one of his bacteria samples in the laboratory during his summer holidays in September 1928, moulds grew on some of them. These had formed substances that killed the bacteria - in this case staphylococci. The first antibiotic was discovered.

Discovery of new pharmaceuticals is becoming increasingly difficult

Such accidental discoveries of medicines hardly ever happen today. It is becoming increasingly difficult and expensive to find new active pharmaceutical ingredients. According to one study, it costs around 2.6 billion US dollars to bring a new active ingredient to market. And before new active ingredients can be used in medicines, they have to undergo extensive and sufficient clinical testing.

But the need for new drugs is there because the growing world population increases the risk of new pathogens spreading. The Corona pandemic clearly demonstrated this. But not only viruses, but also certain bacteria and germs pose an increasing danger: The number of infections and deaths due to antibiotic resistance has risen significantly, as the "European Centre for Disease Prevention and Control" (ECDC) found on behalf of the European Union. According to estimates, an average of 35,000 people die each year in the European Economic Area alone as a result of infections with antibiotic-resistant bacteria.

How can artificial intelligence be used in pharmaceutical research?

For example, in order to discover specific drugs against resistant bacteria, it is important to find new synthetic substances. To develop new pharmaceutical drug candidates, it is estimated that one has to test a decillion molecules. A decillion is a one with 60 zeros, which is a very large data set. AI and machine learning can help to examine this and thus the molecules more quickly. The pharmaceutical company Merck names two advantages of this method, among others:

  • Predictive analytics: Predicting the properties of a substance

    By using AI, it is possible to predict the properties of a substance. This means that only molecules with the desired properties are selected for synthesis. In chemistry, synthesis refers to the process of creating a compound from elements or a complicated new substance from simple compounds. Merck says it has already developed about 300 such AI prediction models for the properties of substances.

    Advantage: Prediction models save time and money because researchers do not even work on substances that have no effect.
  • AI molecule design: Ideas for completely new substances

    AI can also be used to 'invent' new molecules with the desired properties - for example, a molecule that binds a protein involved in a disease pattern, rendering it harmless.

    Advantage: Until now, drug candidates could only be identified through clinical trials. The problem is, however, that only 10 per cent of pharmaceutical drug candidates ever make it to the market after phase 1 studies, i.e. small studies on healthy volunteers. This means that 90 per cent of new active substances undergo extensive clinical testing, but are never used after the test phase. These unsuitable substances can be excluded from the outset by AI and machine learning, thus accelerating the development of effective new drugs.

In which areas of drug development can AI be used?

Artificial intelligence could be used in many areas of the development of new medicines in the future. We have selected two interesting examples from current research.

  1. Antibiotic-resistant infections (Source: MIT)

    Researchers at the Massachusetts Institute of Technology (MIT) and McMaster University in Canada have discovered a new antibiotic against hospital bacteria using an AI algorithm. The bacterium "Acinetobacter baumannii" is often found in hospitals. Anyone infected with it can get pneumonia, meningitis or other serious infections. The problem with hospital germs: they can survive for a long time on surfaces such as doorknobs and are often resistant to all existing antibiotics.

    The researchers identified the new drug from a database of almost 7,000 potential pharmaceutical compounds. To do this, they had trained a machine learning model to evaluate whether a chemical compound inhibits the growth of the hospital bacterium. So far, however, the antibiotic has only been tested in animal experiments on mice. Clinical trials with humans will show whether it will reach the market.
  2. Medicines to treat cancer

    In the field of cancer drugs, the pharma tech company Exscientia is one of the pioneers in the field: In a study with the University of Vienna, the British company tested various potential cancer drugs on a leukaemia patient. They examined dozens of "drug cocktails" simultaneously on tissue samples from the patient and evaluated them with the help of bots and machine learning models. This enabled them to identify a drug that made the cancer disappear two years after starting treatment. However, it remains to be seen whether the promising drug will also successfully pass through clinical trials. The results are still pending. AI may also be used in the development of an mRNA vaccine to fight cancer. mRNA vaccines have been familiar to many of us since the Corona pandemic, and the technique could also be used in cancer therapy in the future. The US company Moderna and the Mainz-based pharmaceutical manufacturer Biontech are both conducting research in the field. Biontech completed the acquisition of British AI company InstaDeep at the end of July 2023 to expand AI-driven drug research in-house.

What are the risks of using AI in drug manufacturing?

If artificial intelligence can find virtually any chemical compound imaginable, it stands to reason that this technology could also be used to create potentially dangerous substances. A group of researchers showed in a study that it took an AI developing drugs less than six hours to invent 40,000 potentially lethal molecules. They put the AI, which is normally used to find helpful drugs, into a kind of "bad guy" mode to show how easily it could be misused to make biochemical weapons. The AI found tens of thousands of new substances, some of which are similar to VX, the most powerful nerve agent ever developed.

Use of AI gives hope, but warning against hype

Given the difficulty of finding new drug substances, the use of AI is certainly an opportunity to find, for example, vaccines against cancer or new antibiotic agents against hospital bacteria. It gives hope that there is currently so much research in the field. At the same time, however, scientists warn against early hype, as some possible substances have so far only been tested in animal experiments or have only gone through phase 1 of a clinical trial. In order to be approved on the German market, drugs must have successfully passed three clinical test phases. In addition, it is important to ensure that such an AI does not fall into the wrong hands, so that an opportunity for humanity does not become a danger.

Most Popular