Drug discovery is a complex and challenging process. Artificial Intelligence (AI) has been used for drug production in the past years, and studies have been carried out in this field. Studies have proved that drug discoveries made with Artificial Intelligence (AI) and Machine Learning (ML) are more efficient and faster than industry standards.
Artificial Intelligence (AI) algorithms occur in many parts of our lives, and studies on the development of Artificial Intelligence (AI) software continue. Among these studies; are topics such as face recognition, eye recognition, the location of autonomous vehicles in traffic, and medical solutions. To ensure success in studies, new algorithms are developed using Big Data, thousands of data are processed, and reflections on the subject continue.
Researchers aim to speed up the process of drug discovery with Machine Learning (ML) and Artificial Intelligence (AI) and minimize the cost. For example, researchers from the University of Cambridge used the developed algorithms to identify new molecules that activate the protein associated with Alzheimer’s disease.
The scenario in making drugs generally includes four different steps:
► Testing molecular compounds,
► Finding the possible effects of drugs,
► Providing a perspective against the disease,
► Using existing drugs and applying them to existing drugs according to test results.
These steps are repeated in the discovery phase, causing a massive workload in terms of both time and cost. While these steps are being carried out, each of the compounds utilized is a candidate for drug development. Some combinations are eliminated during the tests. Work continues with the remaining compounds. Each step in this process is very costly. For this reason, pharmaceutical companies are incorporating Machine Learning (ML) and Artificial Intelligence (AI) technologies into processes to accelerate drug discovery and reduce drug development costs.
From the Cavendish Laboratory in Cambridge, “Machine Learning has made significant progress in areas like data–abundant computer vision,” says Dr. Alpha Lee. With the algorithm developed, Dr. Alpha Lee and his colleagues use mathematics a lot in chemical processes in the stages of drug discovery.
How Are Algorithms Used?
To make the available molecular information usable, researchers first consider the interconnections of each molecule and turn them into graphs with the help of mathematical expressions. Thus, they obtain different data sets with graphical representations. Then, the algorithm looks at the active and inactive parts of the molecules used. For the drug, it determines which of these parts are essential or not.
The algorithm makes predictions about the statistical properties of a noisy dataset, using a mathematical principle (Arbitrary Matrix Theory); It is then compared with statistics of the chemical properties of the active and inactive molecules found in the tests to decide which chemical structures are essential and necessary for research. Thus, molecules that have not been successful in previous studies can be used with this technique. At the same time, Artificial Intelligence (AI) can help test the side effects of candidate drugs.
The researchers built a model that started with 222 active molecules and was able to screen the molecule. “Finding four active molecules out of six million different molecules is like finding a needle in a haystack,” Dr. Alpha Lee said. “The comparability provided by the study shows that our algorithm is twice as efficient as the industry standard,” he said.
We can foresee that Machine Learning (ML) can make the drug discovery process faster and be used in many different fields of medicine. The researchers also share that these studies laid the groundwork for chemists trying to make selections of molecules that would ensure success at the beginning of their studies.
