Russian scientists have improved protein with artificial intelligence

The development will help elderly people maintain their vision

Specialists from the Federal Research Center of Biotechnology of the Russian Academy of Sciences, MIPT and the Faculty of Biology of Moscow State University used artificial intelligence (AI) to modify the AstaP protein (astaxanthin-binding protein). With the help of the machine learning method based on MPNN (message passing neural network), new proteins called NeuroAstaP have been created.

Image source сгенерировано нейросетью DALL•Е 3

AI made it possible to predict the sequence of amino acids in a protein molecule based on the coordinates of the atoms of its "skeleton". Usually, such algorithms work with crystallographic data, but AstaP proteins cannot be crystallized. The only available structural model was obtained by NMR spectroscopy.

In the future, the development will allow people to maintain their vision.

We recently discovered that the antioxidant effect of carotenoids delivered by the natural AstaP protein reduces the harm from lipofuscin pigment granules. This pigment accumulates in cells and becomes one of the causes of age-related retinal degeneration, responsible for approximately half of all cases of blindness.

Nikolai Sluchanko, Head of the Laboratory of Protein-Protein Interactions of the Federal Research Center of Biotechnology of the Russian Academy of Sciences

Researchers have developed a new approach for working with NMR data. Four out of five predicted variants of the NeuroAstaP protein retained the functionality of natural AstaP. At the same time, their sequence coincides with the original protein by only 40%.

Neural networks at the heart of popular algorithms for protein engineering are usually trained on structures obtained using crystallography and electron microscopy. Our proposed approach, based on step-by-step improvement of numerical characteristics of models and computer confirmation of intermediate results, allows us to apply the same algorithms to NMR data without additional retraining, and thereby expands the set of tools for rational engineering of protein molecules for use in medicine and biotechnology.

Ivan Gushchin, Head of the Laboratory of Structural Analysis and Engineering of Membrane Systems of MIPT

Earlier www1.ru reported that a new platform based on artificial intelligence was created in Sochi.