 All right, wonderful. Welcome, everyone, to the session of short talks. The first speaker is Pao, who's going to be talking about D-Couple-R. Take it away. And please hold your questions to the end, and then you can ask questions for all the speakers once they're done. Thank you. OK, thanks. So I'm Paoadie Mumpel. I'm a PhD student at Heidelberg University Hospital in the group of Saez Rodriguez. And yet today I will talk about the coupler, a tool that we have developed that basically consists of a collection of computational methods that infer biological activities in omics data. So in many experimental designs, we have omics data coming from healthy and disease patients. And one field's approach one could take is to use statistical methods to make associations or predictions of the disease. And while this approach is valid, the problem with omics data is that they have high dimensions. So it's hard to interpret. And also we might not have enough statistical power. So in our group, we have specialized in leveraging prior knowledge to extract biological signatures from this omics data. And then we use these signatures to make associations or predictions to disease. This signature extraction lowers the dimensionality of the data, which increases its statistical power and makes interpretation easier. In the group, we fetch prior knowledge from the Omnipath database. Omnipath is a meta resource of more than 100 different primary sources that we have developed in the past in the group. And it contains general biological information such as ligand receptor interactions, functional annotations, gene regulatory networks, pathway networks, and so on. And all this data is available through the R-bioconductor package, and it's also available as a Python package. Another key concept that we use in the group is footprint analysis. Footprint is basically the collection of dumps in molecules of a given regulator. This regulator can be a pathway, a transcription factor, a kinase, or any biological entity that you can link to downstream features. These links are defined by prior knowledge, and they can be coupled to a statistical method to summarize molecular readouts into activities. So here we have a very simple example. It's a small gene regulatory network where we have some transcription factors that target downstream genes with specific connections, edges. And as we can see, we have different values for gene expression, and we observe that NF-kappa-beta seems to have more activities than the other transcription factors because its downstream genes seem to have higher levels of expression, while the others seem to show lower levels of expression. Therefore, they will have lower activity. In this case, for this very simple example, we use weighted sum, but obviously there are better statistical methods to estimate these activities. Some of them are very well known by the bioconductor community, such as gene set enrichment analysis, GSBA, Viper, overrepresentation analysis. And here I forgot to mention AUSL also. The problem with the implementation of these methods is that each one of them uses a specific format, so it's a little bit annoying to switch between methods in the same analysis. So following this idea, we decided to develop the coupler, which is a unified framework with several footprint methods, although other frameworks such as these already exist, like Piano, they are lacking novel footprint methods. And also, inside the coupler, it's very easy to access to all the prior knowledge from Omnipath using one liners. And yeah, this package is available in bioconductor and also as a Python package. While we were developing the coupler, we were also wondering how good are these methods. So we decided to benchmark them using transcription factor perturbation experiments. In these experiments, a specific transcription factor was either knocked out or overexpressed, so we know which specific DF is affected. And using this data and all the methods inside the coupler, we estimated transcription factor activities. And as prior knowledge, we use the gene regulatory network, Dorothea, which is a network that we have developed in the past in the group that gathers DF target interactions following different evidences of data available. So once we have the coupler on the benchmark data set, we obtain transcription factor activities. And since we know which transcription factors are perturbed, we can assess how good are methods at recovering these perturbations, right? So after running the coupler through the benchmark data, the first thing we observed was that all these different methods, they return similar activities, they are correlated. But we observed that they show differences in their predictive performance. We observed that simple models, univariate or multivariate, seem to perform best. And also the consensus score of the methods. And just as an example, this is a study where we successfully used the coupler to analyze this data set. So in this manuscript, we were working with trans-heptomic samples coming from liver from COVID-19 patients. Obviously, we have controls, but from the COVID positive patients, we had two groups, ones that showed presence of the virus in the liver, PCR positive. And the other group that didn't show presence of the virus in the liver. So using the coupler, we estimated pathway and transcription factor activities. And we observed that there was a shift between healthy and PCR positive. And that pathways related to immune response, such as JAXTA, were active in PCR patients. And that collection of transcription factors were changing progressively their activity from healthy to PCR positive. And moreover, this TF signature was reproducible in other hepatic viruses, such as HCB and HIV in other independent data sets. So to sum up, footprint analysis enables the summarization of omics data into interpretable terms. The coupler allows to easily run any footprint method with any resource. And it can infer pathway, transcription factor, kindness activities, and really any biological entity that can be linked to downstream target features. Finally, as an example, footprint analysis identified key transcription factors of hepatic sequel in COVID-19 patients. And yeah, this is it. Thank you for your attention. And thank you for inviting me to this conference. And if you have any questions, do not hesitate to ask. Thanks. Thank you. Very nice talk. Next up, we have Nikolas.