Neural networks in the creation of new medicines

The use of artificial intelligence (AI) in various areas of healthcare is gradually becoming routine practice. Medical institutions have already implemented applications using which the patient can easily make an appointment or see a doctor's prescriptions. AI functionality is already being used in the development of medical devices and medicines. Nevertheless, AI is a fairly broad term that includes a large number of specific tools for solving certain tasks. Today, our article will focus on the use of neural networks to create medicines.

A neural network is a program that can learn based on data and examples. It does not work according to ready-made rules and algorithms, but writes them itself during training. Interestingly, the neural network is designed in the image of the human nervous system, that is, the algorithms in it are arranged like neurons in the human brain. They are interconnected by synapses and can transmit signals to each other.

The neural network learning process consists of the following stages:

1. Getting "input" and "output" data from the dataset (for example, a pair of "text + picture"). At this stage, there is no connection between the neurons.
2. Prediction. For example, the neural network will call a picture of a cat a dog.
3. Error calculation. The neural network determines how accurately it has matched the image and the text.
4. Error correction. At this stage, there is an increase in connections between those neurons that will help recognize the cat.
5. Repeat steps 1-4 until the neural network learns to guess correctly.

Neural networks are already helping to reduce drug development time. In the absence of neural network capabilities, drug development can take up to 15 years, with development costs of up to 3 billion US dollars. The main difficulty lies in the fact that scientists have to study tens of thousands of molecules in order to find the necessary and then determine the optimal ratio of components for the treatment of a particular disease.

Thanks to neural networks, the time spent on creating a drug can be reduced by an average of 10 times. For example, using the GENTTRL neural network, it was possible to find and synthesize a medicine for fibrosis and test it on animals in 46 days. The company Insilico Medicine, which was involved in this study, notes that the drug, developed by artificial intelligence, will be more effective than previously developed ones.

S. A. Chubov, N. A. Tarasyuk THE POSSIBILITIES OF USING ARTIFICIAL INTELLIGENCE IN THE PROCESS OF FORMING THE PROFESSIONAL COMPETENCE OF FUTURE PHARMACISTS