Laadi alla uus Eesti Raadio äpp, kust leiad kõik ERRi raadiojaamad, suure muusikavaliku ja podcastid.

AI delves into microscopic images to help understand how drugs work

The researchers used fluorescence microscopy, which allows them to distinguish details even micrometre in diameter. The image shows the structure of an apple skin.
The researchers used fluorescence microscopy, which allows them to distinguish details even micrometre in diameter. The image shows the structure of an apple skin. Source: Maris-Johanna Tahk and Tõnis Laasfeld

For his doctoral thesis, Tõnis Laasfeld and colleagues analyzed millions of microscopic images to see how drugs work. The results of this research can be used in drug development to help create medicines with no side effects, or particularly long-acting medicines that patients need to take less frequently.

Modern understanding of biology and pharmacology broadly acknowledges that life is a complex dynamic system which has led to widespread incorporation and employment of mathematical sciences such as systems theory, computer science and artificial intelligence in biology and given rise to new fields such as computational biology, systems biology, systems pharmacology, bio- and chemoinformatics and bioimage analysis among others.

In turn, mathematical sciences have borrowed many concepts from biology and natural sciences, an idea known as biomimetics. This has led to the development of global optimization algorithms such as the genetic algorithms and simulated annealing and inspiring the development of neural networks and deep learning.

This shows an exciting and continued interplay and codevelopment of biological and mathematical sciences. In practice, such progress is heavily mediated by various kinds and large volumes of data and software as well as interdisciplinary collaboration helping to turn the data into information and knowledge.

Thanks to these advances in computing and data science, today's drug developers have far more powerful tools at their disposal than they did just a few years ago. As a result, the chances of researchers finding the right drugs from an infinite pool have increased dramatically.

Laasfeld's research focused on applying this kind of interdisciplinary approach more narrowly to develop new tools to study the signaling processes of a class of proteins called G protein-coupled receptors (GPCRs).

GPCRs are an extremely important class of receptors for fundamental biology, being involved in a large portion of all physiological signaling processes. They constitute the most drug-targeted class of all proteins meaning that fundamental discoveries often have a direct impact on drug development.

Tõnis Laasfeld, a recent Ph.D. graduate in chemistry, studied how drugs and their target protein molecules meet and stay together on the cell surface. He combined biochemistry with computer science to create machine learning models and software that analyzes image data. The solution helps to study microscopic images quicker, more accurately and reproducibly.

"Half jokingly, one could say that in the course of this dissertation, a group of nerve cells, or a scientist, investigated how to combine and teach artificial nerve cells, or deep neural networks, to study the functioning of biological nerve cells. On the one hand, this mutual reflection of the natural and artificial worlds suggests that it may not be worth drawing such a clear line between the natural and the artificial; on the other hand, working in both fields is an almost limitless source of inspiration," he said.

It is not only about whether or not a medicine works. It is equally vital to understand how quickly the effect occurs and how frequently the drug must be used. To gather such information, methods for monitoring the binding of a drug to a target over time are necessary, which is what the methods developed in this thesis make possible.

According to Laasfeld, the distinction between old static and modern dynamic approaches is analogous to the distinction between photography and film. The photo just hints at what came before and what might come next, but the film provides a definitive explanation. More information can be gathered by integrating the temporal and spacial dimensions. On the downside side, there is an increasing volume of data from which AI must extract insight.

Such approaches present intriguing possibilities for medication development. It could be especially valuable in the development of long-acting medications to reduce the frequency with which they are administered. Laasfeld and colleagues discovered a prospective medication counterpart that has a very long half-life in the case of the muscarinic acetylcholine receptor, which regulates heart function.

The glowing outer surface is a fatty molecule cell membrane with a drug analogue coupled to receptors with a glowing label. However, inside the cell, a layer of fatty molecules wraps spheres the size of a corona virus, which appear as brilliantly flashing dots. Water spheres are formed when part of the cell's own membrane becomes embedded with the receptor and the glowing drug analogue. The video also shows that many of these spheres move in a rather specific direction, indicating that they are transported by special motor proteins along intracellular "rays" - actin filaments and microtubules.

This medication candidate's molecule could remain linked to the receptor for several days. However, even little changes in the structure of the molecule can eliminate this long-acting property. However, without knowledge of the receptor's function and structure, creating such compounds is extremely difficult.

Another intriguing example is the well-known dopamine receptors, which in the body bind the motivating chemical dopamine. Simply said, when a person achieves a goal they set for themselves, the brain produces the same chemical as a reward. The euphoria that follows inspires the individual to attempt similar undertakings in the future.

Parkinson's disease, on the other hand, and drugs such as cocaine, interfere with the normal release and binding of dopamine to the receptor. When a person develops a drug addiction, the body attempts to compensate for the extra dopamine released by diminishing the number of dopamine receptors. In the short term, the cell retracts the receptor back into the cell, where the released dopamine no longer has the ability to impact the receptor.

The microscopy technology established in this study allows for video observation of how and what happens at the receptor. Contrary to popular assumption, the receptor can transport the medication into the cell. Some chemicals may concentrate inside the cell in the spindle and persist there for far longer than would be predicted from processes that only occur on the cell surface.

These are only a few examples of fascinating occurrences that were previously unknown due to their difficulty in detection. However, the technique and software established in this work already allow for faster discovery and explanation of such discoveries. As a result, even more fascinating discoveries are likely in the near future.

However, the software and procedures developed have a considerably broader range of applications. The novel methods established in the thesis, for example, would allow the efficiency of various cancer medicines to be exhibited on a two-dimensional slide as well as in a more organism-like three-dimensional cell culture.


Follow ERR News on Facebook and Twitter and never miss an update!

Editor: Jaan-Juhan Oidermaa, Kristina Kersa

Hea lugeja, näeme et kasutate vanemat brauseri versiooni või vähelevinud brauserit.

Parema ja terviklikuma kasutajakogemuse tagamiseks soovitame alla laadida uusim versioon mõnest meie toetatud brauserist: