Hunting for new drugs
Pharmaceutical research has been in crisis for years: It is becoming increasingly difficult to find new, effective active ingredients. Many pharmaceutical companies are therefore pinning their hopes on artificial intelligence. But can the self-learning algorithms live up to expectations?
Before an active ingredient is approved for use in humans, it has to undergo numerous tests. It often happens that promising candidates unexpectedly drop out. One of the reasons for this is the cytochrome P450 (CYP450): a series of enzymes that are mainly produced by the liver. They break down various chemicals, preventing them from building up to dangerous levels in the bloodstream. As it turns out, many drugs inhibit the production of CYP450, making them toxic to humans.
Pharmaceutical companies therefore try to find out in advance which candidates could have such a side effect. Among other things, they analyze the potential drug in the test tube, compare how similar, already known active substances react with CYP450 or carry out animal experiments. However, about a third of the predictions made in this way turn out to be wrong. In these cases, only human trials show that the active substance is not suitable - which wastes large sums of money and years of work.
Due to these and other difficulties, drug development in the pharmaceutical industry has been in crisis for at least two decades. Companies are spending more and more money - the ten largest are now investing almost 80 billion euros a year - and are producing less and less successful active ingredients …