Scientists fr om the Laboratory of Protein Engineering at the NSU Department of Natural Sciences are using the Google AlphaFold neural network to predict the structures of repair proteins from the helix-hairpin-helix (HhH) superfamily and to explore the possibility of using them as templates for establishing protein structures by molecular replacement.
The discovery of new repair enzymes will make it possible to solve various biotechnology and fundamental medicine problems. For example, using them to sequence degraded DNA isolated from bone tissue tens of thousands of years old in order to reconstruct the genetic history of mankind from ancient times. They can also be used to analyze ancient burials and archaeological finds to find out at what stage of its development mutations arose in the human genome that cause cancer, when epidemics broke out, and how bacteria and viruses passed from animals to humans.
Dmitry Zharkov, Doctor of Biological Sciences and Laboratory Head, explained,
The AlphaFold neural network is trained on millions of different sequences. Knowing the sequence of the protein and the amino acids in it, we get a model. This can be compared with experimental data to find out the spatial structure of the protein, bypassing the so-called “phase problem”, which makes it difficult to decipher the structure of proteins “from scratch”. Currently, we are engaged in mass prediction of the structures of proteins of interest to us from different types of organisms. At this stage, it is important to understand whether we can use the structures of DNA repair proteins predicted by the neural network to establish their true structure. It may be possible to go a little further, knowing approximately wh ere these proteins have active centers and using the predicted structure to understand how they work. For example, how they look for “broken” DNA links and remove them.
The challenge is that each repair protein has its own specialization and is responsible for the "elimination" of any specific DNA damage. Proteins are combined into groups that are similar in structure, but they have subtle differences in the active site. Scientists try to predict the specificity of a protein based on its predicted structure. But such a prediction has value only when researchers can experimentally verify its veracity. Sometimes the results match, and sometimes they don't. This, according to Zharkov, is the scientific research process.
Knowing the structure of DNA repair proteins will help solve many problems. In particular, antibiotic resistance. It can be overcome by simultaneously acting on several different targets in the bacterium. It is unlikely that it will acquire resistance to several antibiotics at once.
Zharkov provided more details,
We know the mechanism of action for all antibiotics that are currently in clinical use. For example, penicillin and all its "relatives" inhibit the enzymes that bacteria are responsible for building the cell wall. In general, all antibiotics put the bacterium into a state of stress and it tries to defend itself. This stress is accompanied by the production of reactive oxygen species, which cause damage in the cell, including damaging DNA. The combination of an antibiotic and inhibition of the repair enzymes we are working with can significantly increase the sensitivity of bacteria to antibiotics. In fact, a combination therapy is obtained, when one-part acts on a specific function in bacteria, and the other weakens their “defense”. Enzyme inhibitors were also important in the fight against cancer cells as they can be “killed” in a similar way, extinguishing their resistance mechanisms, which makes them more sensitive to therapy.
This research is conducted as part of the large-scale "Structural research and radiation testing of promising materials using synchrotron radiation and neutrons" project that is supported by the "Priority 2030" strategic leadership program. The project team explained that when the SKIF is launched, the method of modeling protein structure predictions using AlphaFold will have one more application, the researchers intend to use the resulting models as starting points for determining the structures of protein c using synchrotron radiation.
