As a biology student, Hovakim Zakaryan developed an interest in stem cells; these are undifferentiated cells with a unique ability to differentiate into a variety of specialized cell types. In the early 2000s, he faced challenges in locating any groups or laboratories dedicated to stem cell research within Armenia's scientific institutions. He eventually discovered a lab at the Institute of Molecular Biology of the National Academy of Sciences of Armenia, which was researching regular cells and viruses. Hovakim developed an interest in viruses while he was working at that lab.
His experience studying viruses at the Institute of Molecular and Clinical Virology in Germany and the Centre for Molecular Biology Severo Ochoa in Spain led to a greater interest in this field of study.
Upon his return to Armenia, Hovakim established the Research Group of Antiviral Defense Mechanisms at the Institute of Molecular Biology, where he and his colleagues began researching the antiviral properties of novel chemical compounds. The research group consisted of three members at the time it was formed. They were looking for antiviral drug candidates against African swine fever virus․
The team expanded over time. In 2020, it achieved the status of a laboratory and was renamed the Laboratory of Antiviral Drug Discovery. The team now consists of eight members, from undergraduates to Ph.Ds.
To gain a better understanding of the research being conducted at the Laboratory of Antiviral Drug Discovery, let's get to know a little about viruses․ Are viruses considered to be a life form or not? This question continues to be a subject of debate among scientists.
While outside of cells viruses lack the characteristic activity of living beings, they do possess genetic material and can multiply when favorable conditions are met, particularly in the host cells of different organisms like humans.
Because viruses are so tiny, they cannot be seen under a standard microscope. Instead, they can be observed through electron microscopes, which magnify their images to a visible scale. Viruses have the simplest structure: genetic material, which is covered by a shell called capsid, which consists of special proteins.
When this parasite with the simplest structure and small size manages to get inside a cell, it starts dividing and multiplying, and its copies come out and infect other cells. As a result, the infected cells either stop acting normally or die. The immune system is also triggered by the virus, and occasionally the immune system's reaction is so aberrant that an inflammatory process begins in the organism. Numerous issues and complications could arise, depending on the virus type and the immune response. From the common flu to fatal immunodeficiency, viruses can cause a wide range of diseases.
The best defense against viruses is vaccination. The purpose of vaccines is prevention. They train our immune system to identify the virus early on and prevent cell infection. However, what should be done if the virus has already infected the cell and a disease has developed? Antiviral drugs come to the rescue here.
Before the development of antiviral drugs, viruses must be studied to identify targets within them and discover chemical compounds that will effectively combat these targets. Since it's a long and intricate process, let's break it down piece by piece.
From the young age of four, Anastasia Shavina was fascinated by surgery videos, which sparked her desire to be a doctor. But when the time came to decide on a career path, she discovered her deep-seated love for chemistry. This led her to choose medical biochemistry.
Over time, Anastasia also became interested in computing. Alongside her university courses, she studied programming in 2020, during the pandemic. Her journey eventually brought her to the Laboratory of Antiviral Drug Discovery, a place where she was able to combine her interests in computers, chemistry, and medicine in one job.
Anastasia's task is to observe the biological processes occurring in virtual reality. Let's explore how.
Finding a target protein in a virus is the first step towards damaging it. Next, you need to find chemical compounds that can bind to that protein affecting the activity of the virus.
Proteins are crucial molecules that drive a wide array of biological functions. Nearly every aspect of your being, from your height and eye color to internal processes like blood circulation and metabolism, is regulated by proteins.
Proteins play a vital role for viruses as well. Proteins on viruses' surfaces enable them to enter cells, where they multiply and spread. Hence, you should target and inhibit these proteins if you intend to disable viruses from functioning.
There is a region on the virus's target protein that looks like a tiny pocket. Chemical compounds can easily get to that region called the binding pocket, settle, and bind to the protein. After binding to the protein, the chemical compounds affect the activity of the virus.
To prove that a chemical compound is effective against the target protein of the virus, the compound must be tested in the laboratory. The cells must first be infected with the virus. After that, the chemical compound must be added to the infected cells. Then you can figure out whether the virus causes fewer cell deaths when the compound is present. Antiviral drugs are developed from the most effective chemical compounds.
Databases with almost all chemical compounds in the world are available. However, these compounds come in billions. Thus, studying each of these compounds in a lab to find out which one is the most effective against a virus would be a time-consuming and expensive process. Therefore, there is a method that reduces the number of compounds studied in the lab.
Using this method, Anastasia selects the most promising chemical compounds from the databases and sends them to the lab. She starts by selecting compounds more likely to bind to the target protein. Anastasia, using the information about the target protein's structure, applies filters to the databases and chooses compounds with the most desirable properties. There may be 2 million compounds left at this stage out of approximately 76 million.
After that, Anastasia chooses the compounds that can effectively bind to the virus protein. For this purpose, she uses software. Initially, Anastasia uploads the target protein's structure into the computer program, followed by a list of compounds extracted from the databases. Based on the uploaded data, the computer program simulates biological processes.
During biological simulation, the software evaluates potential binding sites of various compounds. Once a likely binding site is identified, the program then predicts the type of chemical bond that would be most likely to form between the compound and the protein.
According to Anastasia, the software computes various metrics. However, the key factor is the amount of energy that the program predicts is required for the compound to bind to the protein; typically, the less energy required, the stronger the bond formed. Anastasia selects approximately 100 compounds that have shown the best results based on the program's calculations.
A chemical compound attempts to attach to a protein
However, binding to the protein is not enough, so at the next stage, other details are also taken into account. During this phase, Anastasia inputs data into the computer program regarding the conditions under which the chemical compound would attach to the target protein. For example, the temperature within the cell at the moment the chemical compound binds to the protein is an important piece of data.
Proteins change their shape in response to heat. A well-known example of this is the straightening of curly or wavy hair using the heat from a hair dryer. Hence, the virus protein's binding pocket shape may also change in response to high temperatures. Therefore, the temperature and other details are essential for the computer program to conduct its calculations with maximum precision.
During this phase, Anastasia chooses around 30 of the most promising compounds, which are then synthesized and dispatched to the lab.
Roza Izmailyan meets me at the laboratory. She first gives me the necessary outfit and then invites me inside. With no use of computers, software, or codes, Roza works with real biological processes.
She continues her work even in my presence. While seeding cells for the experiment, Roza engages in conversation with me (of course, the term "seeding cells" is not clear to me, but I'll get back to that later).
Roza received her Ph.D In Taiwan. After many years of employment in Taiwan, she returned home. Following her return to Armenia, she commenced her work in one of the laboratories at the Institute of Molecular Biology. After a while, she made the decision to join the Laboratory of Antiviral Drug Discovery, as its research area was a better fit for her interests.
Roza and I begin our lab journey by getting to know cells. Roza first needs cells extracted from organs for the experiments. These cells are maintained in an environment that provides the necessary nutrients to keep them alive.
She first puts, or as the formal term says, “seeds” the cells in a plate designed for this purpose, before infecting them with a virus. After that, she infects the cells with a virus and waits for some time. Following that, Roza adds the chemical compounds to the infected samples.
Cell deaths are observed in order to assess a chemical compound's efficacy. Roza starts by calculating the number of cell deaths due to the virus, and afterward, she checks if the presence of a chemical compound contributes to a reduction in the number of dead cells. The decrease in cell deaths indicates that the compound is affecting the virus.
Arthur Ghazaryan walks in while Roza is telling me about the work being done in the lab. Artur is a student at the Medical Biochemistry Faculty. He has always had a passion for research, and that's what led him here. Arthur started out by simply watching the scientists at work in the lab, but gradually, he began to participate actively in the tasks as well.
Roza explains that to be part of their lab, there are two key requirements: first, a deep love for the work, and second, harmony between one's mind and hands. Arthur, meeting these requirements, was thus invited to join the team.
Roza picks up where we left off after introducing me to Arthur. She explains their method of estimating the number of cell deaths. With a microscope, you could hardly count the number of cells that were killed by, say, 5 milliliters of virus. This is because the number of dead cells is very large making it impossible to observe each death under a microscope.
This is when the dilution method emerges as a savior. In this method, the virus is diluted several times using a chosen solvent. The concentration of the virus begins at 1% of the solution, then diminishes to 0.1%, later drops to 0.01%, and keeps decreasing in this manner․
The diluted virus is then used to infect new samples, and since there are fewer dead cells, it is then possible to see each death under a microscope. If a solution with a 0.01% virus concentration leads to the death of 10 cells, it's easy to calculate the number of cells an undiluted virus would kill.
Roza and colleagues use another method called cell staining. Using this lab technique, cells are colored to make them more visible under a microscope. In the infected samples, scientists add a purple dye capable of staining only the living cells, not affecting the dead ones.
The scientists stain the experiment samples and then observe the result. A greater number of unstained regions indicates a higher number of cell deaths. Cell staining is performed at the concluding phase of laboratory experiments to offer a qualitative assessment of the research conducted.
Rafayela Grigoryan walks in as Roza is showing me their stained samples in the lab. When Rafayela first heard about the Laboratory of Antiviral Drug Discovery, she was a biology student.
She wrote her master's thesis in this lab, which was later published as a scientific paper. Rafayela now is a part of the team. She is also a graduate student of the Institute of Molecular Biology, where her dissertation is supervised by Hovakim Zakaryan.
At the Laboratory of Antiviral Drug Discovery, all experiments are in vitro, taking place outside of living organisms. Following that, the drug candidates will need to go through in vivo (involving animals) and clinical (involving humans) experiments. The most effective drug candidate, having passed through all the required stages, will ultimately be produced as a drug.
In 2019 Hovakim Zakaryan, the Head of the Laboratory of Antiviral Drug Discovery, and data scientists Mher Matevosyan and Vardan Harutyunyan came up with a startup idea. The idea was to apply Machine Learning to design new antiviral drugs․
In the year 2020, they successfully founded their startup, Denovo Sciences. Denovo Sciences runs separately from the Laboratory of Antiviral Drug Discovery, but it collaborates with the lab.
Machine Learning (ML) is a branch of Artificial Intelligence. In ML, algorithms or statistical models are developed, which make predictions after being trained on large-scale data. In the field of drug design, the use of ML has significantly increased in the past few years. The use of generative methods is standard practice for this purpose.
Vardan explains how the generative methods work. Imagine you aim to develop an ML model capable of generating human images. For this task, you will input a large dataset of human images into the model for training. After being trained using this dataset the model will become capable of generating human images by itself.
If your goal is to create a model capable of generating molecules, you would follow the same steps. Your initial step will be to input a large selection of chemical compounds into the model, followed by offering examples of compounds that have efficacy against a target protein. After processing the data, the ML model will generate novel molecules it predicts could effectively bind to the protein.
The accuracy of the data used for training the model is crucial. Inaccurate or low-quality data can cause the model to make faulty predictions. Hovakim states that the data available on viral diseases is both insufficient and of low quality. That's why Denoco Sciences chose Reinforcement Learning instead of generative methods. Reinforcement Learning works well in situations where there is a lack of data or low-quality data. This method enables the training of an ML model without relying on large datasets.
According to Mher, the most famous application of Reinforcement Learning is the AlphaZero program created by the DeepMind․ AlphaZero quickly learned chess and demonstrated impressive results. Mher explains how DeepMind managed to create such a program. Initially, the company adopted a generative method, training its model using games played by the world's top grandmasters.
Afterward, DeepMind turned to the use of Reinforcement Learning. For this attempt, only the information regarding chess rules was provided to the program. AlphaZero began playing chess against itself, while the DeepMind team provided feedback on the quality of its gameplay. The program reached a high level of proficiency in chess, doing so without relying on the game's human-based knowledge.
Using the same method Denovo Sciences developed an ML model that is designed to generate molecules. Similar to how DeepMind inputted chess rules into its program, the startup team provided its model with the principles of chemistry, biology, and physics. After that, they gave the model the desired protein's structure and let it “play the game”:
Around the target protein, the Denovo Sciences model started generating novel molecules. It seemed like all of the puzzle pieces were assembling around a single, tiny piece. Throughout this process, the Denovo Sciences team provided the model with feedback on the efficacy of the molecules it generated. Ultimately, they developed an algorithm that, as Hovakim states, generates molecules with the best possible properties.
New molecules generated by the Denovo Sciences model are synthesized and distributed to different labs. The Laboratory of Antiviral Drug Discovery studies compounds against the influenza virus. In partnership with the Center International de Recherche en Infectiologie (CIRI), Denovo Sciences is focused on creating a universal drug to combat all coronaviruses. Recently, Denovo Sciences entered into a partnership with Singapore's A*STAR ID Labs aiming to jointly discover drugs to combat the Dengue virus.
Right now, Denovo Sciences is dedicated to designing drugs aimed at combating viral diseases, although their model has the potential to generate molecules for various targets.
"Should there be a prospective collaborator with an interest in cancer or another disease and possessing a target, we are open to partnership," says Hovakim.
Recognizing the shortage of professionals in this field in Armenia, Denovo Sciences is committed to the education and training of new specialists. Mher Matevosyan and Hamlet Khachatryan, members of the team, are engaged in lecturing on chemoinformatics at the American University of Armenia.
It's noteworthy that Denovo Sciences also contributes to new knowledge. Impressively, the team has succeeded in publishing a scientific paper. Another of their articles is presently in the process of peer review.
Author: Anna Sahakyan, video by Roman Abovyan, photos by Roman Abovyan, Sargis Kharazyan, and Julietta Hovhannisyan
This article was prepared with funding from the Young Scientists Support Program