The process of making new drugs is getting more difficult in 2023. In the future, AI could help pharmaceutical companies get new products on the market more quickly.
Alexander Fleming found penicillin, one of the most important antibiotics, when the medium in which he grew one of his bacterial cultures grew mold. Such accidental discoveries are rare today. Drug development costs billions of dollars, which the pharmaceutical industry has to spend. Still, about 90% of active ingredients fail during clinical testing, which is the last step in development.
At the same time, we need more and more new medicines. This is because the world’s population is growing, which makes it more likely that new diseases will appear and spread. Since the beginning of the 20th century, flu, corona, and HIV pandemics have killed up to 100 million people. At the same time, bacteria are always coming up with new ways to resist antibiotics. With that being said, we need a way to utilize AI in helping pharma companies.
AI can significantly shorten the development process
Researchers can use data on the effectiveness, bioavailability, and side effects of known drugs to teach artificial intelligence how to predict these things in newly discovered drugs.
New active ingredients often fail because it turns out that they stop the liver from making cytochrome P450 (CYP450), which is an important enzyme. These are important for the liver to process substances that don’t dissolve in water. They also help make hormones and other substances.
Since it is not always clear why cytochrome production is slowed down, it has only been possible to predict with a success rate of 60–70% whether an active substance slows down CYP450 production. With artificial intelligence, the rate could go up to 95%. Because of this, there are a lot fewer active ingredients that need to go through preclinical and clinical testing. This means that in the future, it will be easier and faster to make drugs with fewer side effects.
AI can significantly shorten the development process
Researchers use data on the effectiveness, bioavailability, and side effects of known drugs to teach artificial intelligence how to predict these things in newly discovered drugs.
New active ingredients often fail because it turns out that they stop the liver from making cytochrome P450 (CYP450), which is an important enzyme. These are important for the liver to process substances that don’t dissolve in water. They also help make hormones and other substances.
Since it is not always clear why cytochrome production is slowed down, it has only been possible to predict with a success rate of 60–70% whether an active substance slows down CYP450 production. With artificial intelligence, the rate could go up to 95%. Because of this, there are a lot fewer active ingredients that need to go through preclinical and clinical testing. This means that in the future, it will be easier and faster to make drugs with fewer side effects.
Hundreds of thousands of potential medicines
Many active ingredients have been found in nature, and more are being found all the time. But this reservoir will run out of water at some point. Also, synthetic active ingredients can often have a much more targeted effect on a particular pathogen and can get around resistance. The number of molecules that could be used as active ingredients is thought to be around 10100.
A decillion is a number with 60 zeros after it. AI can’t study all of these molecules, but it can study more of them faster than humans ever could.
KI-Molecule-Design
AI can also help scientists find these molecules in the first place. Since there are so many, it doesn’t make much sense to test each molecule one by one to see if it could cause a disease. For example, if it is known that a certain protein is involved in a disease, AI can be used to create a molecule that binds to this protein and makes it harmless.
AI is also used to come up with new cancer treatments based on mRNA. In December, the US company Moderna published a clinical study in which an AI chose the surface proteins of skin cancer cells, which were then vaccinated against using mRNA. BioNTech, a drug company based in Mainz, Germany, said on January 10 that it would buy the British AI company InstaDeep in order to use AI to develop more drugs in-house. Both AI and mRNA could be used to make medicine more personalized in the future.
AI can also develop bioweapons
Instead of making drugs, the US company Collaborations Pharmaceuticals tried to use its drug AI to make molecules that were as dangerous as possible. Without knowing what it was, the AI made things like VX, which is a very powerful neurotoxin. The results of this experiment were later destroyed, but one thing is clear: in the wrong hands, this kind of tool can also do damage.
Still music from the future
No one knows for sure how well AI will work in the pharmaceutical industry. Clinical trials are already being done on some active ingredients that were made with the help of AI. Exscientia, a British company, is testing a few of these substances, like the cancer drug EXS-21546.
But it’s still not clear if these substances are better and have fewer side effects than drugs made in the traditional way. It is also not clear if the development is quicker and cheaper. AI is just starting to be used in the pharmaceutical industry, and it will be a while before it can keep up with traditional methods.