It turns out that it is very difficult to answer a rather simple and almost childish question: “Why does it take so long to make new drugs?”. Most often, the development of a new drug takes more than ten years, and the cost of this development can reach billions of dollars. This is due to the fact that about 90% of experimental drugs fail at various stages of creation and implementation, i.e. while formulating and testing on animals or humans. Drug development is a complex process Artificial intelligence and drug development How does it work? Drug manufacturers and investors have begun pouring billions of dollars into artificial intelligence-powered discovery systems for new treatments. Scientists are striving to discover cutting-edge medicines by quickly identifying new compounds and simulating complex mechanisms in the body, as well as automating as much as possible what was previously done manually. So far, only a few drugs based on the highly publicized technology have reached the testing stage. Classically, machine learning occurs using preloaded data and special algorithms. And artificial intelligence allows you to simulate these processes, providing intuitive human interaction with this algorithm. Drug development is a complicated process There are about 10,000 diseases that affect people (most of them are quite rare), and for many of them there is no effective treatment. Often, scientists can only guess at the mechanism that causes a disease, let alone determine a treatment or cure. The process of choosing a therapy from the many trillions of possible drugs synthesized chemically or from biologics, usually produced using genetically modified cells, is also largely based on pure guesswork. Because of this, the development process is closely related to the principle of trial and error. Artificial intelligence and drug development Pharmaceutical companies are joining forces with technology divisions to process huge amounts of data. Some of these have only recently become available and have been digitized with the advent of electronic medical records. The main goal is to help identify promising treatments. To do this, companies use deep learning, a type of machine learning in which neural networks consisting of many layers are trained on large data sets to learn to recognize and structure patterns and make predictions. It is still too early to tell whether AI will lead to the creation of new life-saving drugs. However, it can at least help identify ineffective drugs earlier, before companies spend a lot of money developing them. For example, AI can better predict whether experimental drugs will have a fairly common, often horrifying side effect that inhibits the production of drug-breaking enzymes in the body or prevents them from being produced to toxic levels. How it works? Here are some examples of models. Identification of proteins that may play a key role in the disease. The AI can sort through datasets, such as data from tissues or blood samples from patients who have the disease and from other patients. This could help researchers focus on developing new compounds that can specifically target these proteins. Robots can apply thousands of potential drugs to different types of diseased cells in hundreds of thousands of miniature experiments. AI can be used to process millions of images of cells - more than a person can view on their own in a lifetime - to predict whether experimental treatments can prevent or stop disease without harming healthy cells. Processing data on existing drugs can uncover ideas for changing prescriptions, combine them to make them more effective, or find ways to personalize them based on genetic markers. Swiss drugmaker Roche Holding AG aims to use AI to double the number of drugs it can produce within 10 years. GlaxoSmithKline Plc hopes to double its new drug success rate to 20% through the use of AI by teaming up with partners that include DNA test provider 23andMe Holding Co. and Cerebras Systems Inc., a new chip manufacturer that provides computer systems for handling very large datasets. Machine learning algorithms can be used at all stages of the drug development process. For example, to find new uses for them, predict drug-protein interactions, determine drug efficacy, provide safety biomarkers, and optimize the bioactivity of molecules. Machine learning methods are gradually penetrating and taking root in mathematics, N&T wrote about the recent application of AI in representation theory and topology. BrainIn principle, a person cannot imagine and realize multidimensional objects in full: we can work with spaces of dimensions higher than the fourth, only using intuition from working with two-dimensional and three-dimensional spaces. Due to the limitation of our perception of multidimensional objects, we cannot effectively analyze them and see all the relationships. This is where machine learning helps: in searching for patterns in huge data sets, the dimension for it is not the limit.