 So the next speaker is Professor Francisco Planes and he is principal investigator at the University of Nevada in Spain, focusing on the development of algorithms for molecular network analysis in cancer and personalized nutrition. And he will talk about the science underneath the IDIOT app. Yes. Oh, sorry. One more shift, please. OK. OK. Thank you very much, Darius. I'm going to talk about the algorithm to rank food according to gut microbiota. Sorry. Our starting point is that the human gut microbiota degrades input dietary compounds into key microbial metabolites that regulate host health. We have metabolites that are health promoters and we have metabolites that are disease promoters. That's our starting point. The question is, can we control through diet the production of these key output microbial metabolites considering that each individual has a different gut microbiota composition? That was our question, how to control the production of these key output microbial metabolites? But the previous fundamental requirement and scientific question is that if we know how input dietary compounds are degraded into key auto microbial metabolites, can we know, can we model, how these input dietary compounds are degraded into these auto microbial metabolites? So to answer this question, our approach was to rely on metabolic networks. At that time, it has been released in Nature Biotech, an impressive repository of metabolic reconstruction of almost 800 members of microorganisms of the human gut microbiota, developed by the group of INES-TLE and the repository is called Agora and it can be extracted from this web page. Networks integrates thousands of reactions that occurs in different microorganisms and they are stored in different databases. We were going to work with this one. With these networks, we can build microbial community models where different species can be treated as a different compartment. And what is most important with these networks, we can connect input dietary compounds with output microbial metabolites. It's similar to logistic network, to city networks, but here we have reaction network where different denotes or the city are metabolites. I'm not going to put any mathematical formula. But what is even more interesting is that this community model can be contextualized for different individuals using metagenomic data, in our case, cystiness, RNA gene sequencing data, taken from stool samples. So depending on the microorganisms you have in the microbiota, you're going to have a different metabolic capabilities. In addition, in this community model we can contextualize the different foods we are taking because different foods has different input compounds here. And again, we're going to have a different output microbial metabolites. We're going to have a different metabolic capabilities. So to build this context specific or personalized community models, we integrate different data from different partners in the instance for health. For 16S RNA sequencing data, we have here Pilar. And also all the information about foods with the different databases generated in the consortium, also with the help with Professor José Ángel Rufián and EDIET. So we have all the ingredients to build this personalized microbial community model. Our work was to, once we have all this data, take all this data and predict for different foods the output microbial metabolites that are going to be synthesized. And we start to do this work. But we found that in Agora, in this repository, was the information about key input dietary compounds was insufficient, was difficult to connect the information we have with EDIET with Agora because many degradation pathways for these nutrients was not annotated. So we developed a different complex by informatics and combination pipelines to complete these networks that was more amenable to integrate with EDIET. We validate with different data sets and different nutrient families that we were more accurate than Agora to predict the output microbial metabolites using targeted metabolomics data, but also uses untargeted metabolomics data. And we did very different compounds nutrient families. So now this was an important milestone in the project. We have the human, a more accurate human metabolic network of the GATT microbial era. But the question is still there. How Agreda, our repository, our human metabolic network is called Agreda. It's an improvement of Agora. So Agreda actually defined how input dietary compounds are degraded. But the question is still there. How can we use Agreda to personalize the diet and rank food for its patients? So we follow this strategy. We select a number of target metabolites based on literature and metabolomics data. Some of them were healthy, and some of them were unhealthy. That was our first step. Define this list of auto microbial metabolites we want to target. We want 42 metabolites we want to follow, and they are going to give us the states of the human of the GATT microbial era. And then, using our computational approach, using the metagenomic data, and Agreda, we rank foods according to a global school that considers the capacity to produce more of healthy target metabolites and less of unhealthy target metabolites. Here, we have four metabolites. And we analyze for each individual the capacity to produce more of the healthy. These are healthy, sort chain fatty acids, and one here is unhealthy. And we construct a normalized integrated score for the capacity of different foods for different individuals to produce these metabolites. This is interesting to note that this curve here was built using a reference distribution from healthy individuals taken from a completely different cohort. So somehow, we can give a normalized score considering healthy individuals of the capacity of the different foods to produce more of the different target metabolites. Here, we have here four foods, four different samples. Obviously, we can find some of the food that are always, in general, are good for the microbiota. And we also have other foods that are general bad for the microbiota. This is Spanish chorizo. OK. This is broad beans. We have we I took a lot of broad beans in the lunch. This is a secret. But also, the interesting thing was this one, this chestnut, that for some individuals are very good and for some individuals are not that good. And this is very dependent on the gut microbiota, OK? Possibly, these individuals lack some particular microorganisms that are making a much better use of these different foods. OK, this was applied, this approach was applied to all the samples and different interventions we have in the projects. Yes, as a final concussions, using agreda, metagenomic data and food databases, we have built a novel ranking algorithm, food ranking algorithm for each individual, according to gut microbiota. This is something very unique in the scientific literature. The second point is that the definition of target metabolite is a central decision in our approach that can be adapted to specific clinical conditions. And this is something we can continue improving for different versions of the app. Because perhaps for diabetes patients, we have a different panel of biomarkers, of metabolite biomarkers. We are also now conducting additional experimental validation to demonstrate this ranking of food using in vitro fermentation experiment that were generated in stands for health. And as I just saw, we have also integrated our microbiota-based food score with iodide soap in order to provide individual nutritional recommendation in human intervention. So thank you very much for all the technical, my PhD student. Telmoam Francesco, in the frame of this project, has conducted their PhD studies. And thank you, the commission and all the partners stand for health. Thank you very much. Thank you. We have five minutes for questions. It's Lidl. Yes. Is this all the questions? Are these from before? Well, that was quick. Better only come here and then I'm going to go. OK. Yes. So maybe the most important question for some, is Agreda open source somewhere? The human metabolic network, Agreda, is open source and is published in nature communication. So it's already available since 2021. What is not available is the ranking algorithm. We're still deciding what to do. Possibly we want to protect it. But yes, the Agreda is open source. If Anonymous was interested in, if they could get access to the algorithm, is that something in the future? I have to, we cannot answer these questions right now. We are also studying, if we are going to publish the algorithm, we are going to explode it. We are in that discussion. So currently, we have an algorithm available for the public, really available, obviously. Other questions? Yes, in the back. We'll get you a microphone. Need to find my way first here through. Yeah, thank you for a very nice presentation and also the very interesting information you have got from assembling all this data. But I was wondering, of course, everybody is using fecal samples to monitor the microbiota, my metagenomics. But I was wondering, don't we miss something? Because the digestion of food, of course, is happening in a small intestine. And what we see in the fecal samples, of course, is not resembling the small intestinal microflora. Of course, it's very difficult, or even not possible, to sample the small intestine. But I was wondering, aren't we missing important information by just focusing on fecal samples and so what's happening in the large intestine, the colon? Yeah, I agree. This could be an additional improvement in our algorithm to not only take and consider the microbiota, but also consider that these outputs could be also degraded in the intestine by intestinal cells and produce another and get to the blood with another transformation that could be mediated. But in the project, we have to, as Pilar says, we also have boundaries in the project, and we decided to stop and validate with the gut microbiota. That's I agree with you. We could improve the outcome of the algorithm by considering also the metabolic contribution of intestinal cells. For example, one additional limitation of our algorithm we are currently working on, we are improving, is that at the moment in the algorithms, we use the microbiota composition as something static. You have a photograph of your microbiota when you take the sample, but this is something very dynamic. And you have also interactions, nutrient interaction with bacteria. And we have created a new version of the algorithm considering this dynamic interaction between bacteria and nutrient. And this is something, in many cases, we have interaction that are not metabolic, but also regulatory. And this is something very interesting that, but yes, we have many ways to improve the, and hopefully we can, we have the opportunity to keep improving our approach in the future. Next we have questions from the online audience. And one from Stephanie Bodenbach, asking if the screenshot sample recipes refer to, for example, Serrano Ham, do you advertise geographic indication covered foods in the app? Is this appropriate for EU-funded project? I'm not sure if this goes to the algorithm itself, or if you can speak to that. Yeah, I think we have three. But if not now, maybe this question will be answered then after the conference. And the next one also goes more to the app. So, well, thank you, Professor Francisco Planes.