Skip to main content

Receptor discovered using artificial intelligence model developed by SZBK opens up new ways to protect against coronavirus

Highlighted News

Péter Horváth, director of the Biomag Research Group of the Szeged Biological Research Centre (SZBK) at the Eötvös Loránd Research Network and Head of the Biomag Research Group, and his partners, Peter Cullen and Yohei Yamauchi, research professors at the University of Bristol, have shown that the SARS-CoV-2 coronavirus can enter the host cell through a hitherto unknown actor, the neuropilin-1 (NRP1) receptor on the host cell surface, which they discovered through their research on influenza.


Many research laboratories around the world are working to help develop effective treatment by understanding the process of coronavirus (COVID-19) infection. Researchers have so far been able to identify the angiotensin-converting enzyme 2 (ACE2), through which the virus is able to enter cells. Research based on SZBK's artificial intelligence model suggests that NRP1, in addition to the already well-known ACE2, may be a new, second focal point for the treatment of COVID-19.

Neuropilin-1 (NRP1) is a receptor found on the surface of a host cell to which the SARS-CoV-2 virus is able to bind through a protein called S (Spike). From this S protein, the S1 protein is formed by enzymatic cleavage, which has a special pattern, the ‘C-end rule’ (CendR), at one end, the so-called C-terminal end. Through this region of the protein, the virus is able to bind to NRP1 and enter the cell. Infected cells, unlike healthy cells, have several nuclei.

To detect and quantify this difference, Péter Horváth and his team have developed a method that is unique in the world and based on a new trend in artificial intelligence, deep learning that enables researchers to perform very accurate microscopic analysis.

Previously, the Szeged research group used a similar methodology to screen for the NRP1 gene in influenza research. They gave this algorithm the name nucleAIzer ( Intelligent algorithms, such as those used to control self-driving cars or for intelligent analysis of images on social media, need huge training databases that the research team did not previously have access to. They therefore developed a hybrid method employing a deep learning method to generate artificial examples and train another intelligent method based on them. The method has just been published in the most prestigeous journal of systems biology, Cell Systems (

The accuracy of the algorithm is shown by the fact that it allowed the Biomag Research Group in Szeged to achieve the highest score in one of the largest global competitions.