Researchers from the Polytechnic University of Madrid and other international centers combined mouse experiments and machine learning models in the laboratory to analyze protein effectors, which enable pathogenic bacteria to evade the immune system and its response.
Many disease-causing bacteria use molecular “syringes” to inject their various proteins, called Effector, In the intestinal cells, thereby blocking key immune responses.
Now, an international team of scientists from the United Kingdom, Israel and Spain, the Polytechnic University of Madrid (UPM) has joined the team. They combined laboratory experiments and research tools to analyze all these protein molecules together. Artificial Intelligence (AI).
Authors, they publish their findings in a journal science, They have used 100 kinds of bacteria Citrobacter rodents From mouse Simulate the functions of all effectors in the body. In this way, they found that they can work together as a network, allowing microbes to have great flexibility, bypass the immune system and maintain their pathogenicity.
The artificial intelligence platform can correctly predict the results of alternative network colonization based on in vivo data.UPM Researcher, Professor of Artificial Intelligence Alfonso Rodriguez Patton And PhD students Elena Núñez Berrueco Use data collected in the laboratory to build Machine learning model.
Novel AI technology
The number of possible combinations of effectors exceeds one billion, so studying all variants will require more than a thousand years of experimental research. That’s where AI comes into play. It can change the rules and allow decryption of this complex mechanism. After learning the model of 100 laboratory experiments, the algorithm developed by UPM can predict the infection ability of any variant.
Núñez explained: “By studying such a complex biological system, artificial intelligence can see what we can’t see.” Prediction helps us determine the most relevant combination of effectors, thus saving time and resources.We can use this model to Predict whether there are new strainsAnd combined with effectors that are different from the research object, Can manipulate our cells And the way he does things. “
This algorithm It is inspired by artificial neural networks, but combines knowledge about effector targets. The architecture of the network is unique: it is not universal, but has the same shape as the biological interaction network of effectors and cellular components. In this way, the network can be trained in very few cases and can also produce models with interpretable results (so-called interpretable AI).
With the help of this model, scientists have been able to lead the following experiments to the most interesting variants.So they can find These essential small molecules. This means that when bacteria are removed or enclosed, they will not be infected, setting a promising target for future treatments to help defeat these skilled invaders.
In fact, the authors have also observed that host mice have the ability to adapt to overcome obstacles erected by different effector networks and activate complementary immune responses that eliminate pathogens and induce protective immunity.
Rodríguez-Patón concluded: “Artificial intelligence once again Disruptive technology, In this case the field of microbiology. This interdisciplinary research requires us to develop novel AI technologies to reveal the complex network of molecular signals that bacteria use to infect us.The results obtained are very satisfactory, so we will continue to work with Gad Frankel -One of the main authors of future research at Imperial University London”.
David Ruano-Gallego, Gad Frankel, etc. “The deconstruction of the effector network of type III secretion system reveals the inherent robustness and plasticity of pathogenesis and immunity.” science, 2021.