 Okay, so hey everyone, so we are going to start, so we are going to start module six, which is called function prediction. So, so the module was linked to a pre-recorded, sorry, a pre-recorded video for Queen Maurice that I hope you had the opportunity to view during your pre-work. And so we are going to use a software called GeneMania as our functional prediction tool. This tool has been developed a few years ago by Craig Morris in collaboration with Gary Bader and it is still maintained in the beta lab at the Donnelly Center. So these slides are just a summary. I don't want to be redundant with the lecture that you viewed, but I know that some concepts used may be a little bit complicated, so that's why I wanted to review in a few slides, but also wanted to show you as examples as we use GeneMania in our projects. So GeneMania is a functional prediction tool designed to answer two questions. So the first one, I have one gene in my query and I want to predict the function of this gene, so the question is what does my gene do? And the other questions, so as a query we have a genelist, so we input a genelist in GeneMania, and the question that we would like to answer is give me more genes like those one. So more genes like the one in my genelist. So we want to expand the genelist. So here in a very short summary, here are some keywords defining GeneMania. So GeneMania is a functional interaction network. Nodes in the network are genes, proteins, and they are connected by edges if they are functionally associated. So for example, the proteins may be known to physically interact with each other or to belong to the same biological process or known to colloquialize. All function networks are merged into one global network, and there are weights associated with them. And those networks help to predict the function of the genes and to add to relate new genes to the network. And GeneMania is used as the concept of built by association to find a gene and networks associated with our genes of interest. Finally, GeneMania is available as a web app or as a set of skip app. And we are going to show you both options today. So this slide is a summary of some of the GeneMania concepts that we have seen in the previous page and in the recorded video. So like the pencil piece are representing the different network. So we can have a network representing the knowledge that we have about protein-protein interaction domain or shared protein domain or genetic interaction, colloquialization, co-expression or pathway. And each of this network have weight and we combine all of this network into a global network. And we have the edges represent these different networks and are colored by different colors depending on the network. So label propagation can be used to associate a gene. So here is my gene, query gene. So I would say at the beginning it's white because it has no function. And we want to associate this query gene with our existing network. And we are going to use a label propagation and the built by association. So this query gene is not associated with this little cluster but is associated with this big cluster by one edge which is I think physical interaction is red. So we know that this query gene is known to physically interact with this one. But we don't know the function of this query gene. However, this big red nodes here for those genes, the function is very well known. And we can see that my query gene is then related to these genes with a very known function by a lot of interaction. And by label propagation we are going to guess the function of this new gene. So context dependent network, I think that there is like one aspect that is sometimes difficult to understand is that there are different weights for the network to use to measure node connections. And there is an option in gene media you can use network weighting or automatic or network weighting equal by network. So in order to understand that we are going to try it in the lab. And so by default gene media is going to use automatic weighting and is going to focus on the biological pathway. So this is the blue. I think it's the blue. I don't see very well. Like a light blue color or greenish like blue color to weight the network because the goal is to predict the function and the function is very defined using the biological process go BEP from gene ontology database. But if you want to have more information about all the networks like physical interaction, pre predicted interaction and so on, you can set the option to equal by network. And then you see the percentages. So that's the percentages that were used to build the network. And so here the pathway was six point 17% and reduced to 3.05% when we use the equal by network option. So then we are going to solve the examples on how to use it in our project. And so as I said, we can use gmini for to query one gene and to to guess the function of this gene and to a gene list. So for one gene, so, so all those examples are from my own project. So I was working with the gene ipo four. And I didn't know much about this gene because I was just starting the project. So I just entered ipo four into gene media. And this is the network that had obtained. So it gave me for example, ipo five, which is working with ipo four. And most importantly, it gave me the function predicted by ipo four. And it's poor complex and nuclear poor complex. And indeed, ipo four is an important and it works. There may be about 10 importing in the cells and it works especially with ipo five. So it works in to import protein in the nucleus of the cells. So, so they bind cargos like protein that are in the cytoplasm. And the role is to import them in the nucleus. So I think that you may guess the function pretty accurately. So on the next tree, I'm going to input two genes. And here, so I was working with this two genes, Zeus 12, and I don't know how to pronounce Zeus 12 and is that H one. And I knew, I knew that those genes were part of the PRC two complex. So the polycomb repressive complex two. This is a chromatin associated method transfer transfer. So it's a it's a group of cells, it's a group of proteins involved in repression. And my goal here was not to guess the function because I knew the function and it was here predicted by gene mania. But my goal here was to retrieve all the other protein part of this complex. And I wanted to do it in a quick way. So what I did, I just input Zeus 12 and is that H one into gene mania. And then I retrieve the genes related to that. And this is the image of the PRC two complex. And you can compare both of them. And then you can see that the main parts, the main protein partners of the complex, were indeed retrieved by by gene mania. Then you can see here, is it H one, Zeus 12. I think dry two is there and I'll be P four and seven are there. And I think this one is there as well. So what I did is just extract this network to to further apply my analysis. And then so then and we expand our journalists. So now we are going to create 23 journalists. So all of them one query to query. Before that, I was doing this on the gene mania web app, because it's query that are easy to do. And this one is also on the gene mania app. But if my query becomes a bigger and bigger than a switch to cytoscape. So by the design actually is the app. So I need to modify this. So that's the cytoscape web. So I have 23 genes. And those 23 genes, three genes, they were genes mutated in Alzheimer disease. So that was SNPs and indels in genes from Alzheimer patients compared to controlled cases. So 23 genes. So those genes, I didn't know about that function. And because it was not arenas, it was not transcriptomics, I did not expect them to be related in a pathway. But I didn't know I was hoping that they were related. I think it's from paper. It's not my project. So I just put them in gene mania. And I was surprised to see connection. And that's what I was hoping for. So I could see really some non physical interaction between those genes. Although the pink. And the function also was really correlated to neurons. So that was good to know that those genes that were just picked up because they were frequently mutated in Alzheimer disease could connect with each other and predict this function. And so now we have 43 genes. So this is done in cytoscape app because we then we then can color the network with different colors. So this is a project that was done a few years ago in Gary Beger's lab. And what they had, they had 43 genes involved in cholesterol metabolism related genes. So the genes that they had were the large nodes. And they used gene mania for three reasons. First, they used gene mania to get some linker genes. So they had some linker genes. For example, those one are very interesting because they link those genes that they had in their in their analysis to other that were not connected. So they increased the network by adding linker genes. And now they could also see that by these golden ages that many of the genes were sharing protein domains were related by protein domains. And the third, the third layers and this is one good aspect of creating a network with cytoscape. So they could overlay that with additional additional information, which were time points. So red were up regulation in their study, blue were down regulation in their study. And they could put as a not center. This is an older version of enrichment maps. So gene mania. So they could color the not center by the time by three hours and the not border by the time point 21 hours. And then the last, the last slide should sound should look like a bit familiar to you because this is how we move from enrichment map to gene mania. So now for sure we are doing this in the cytoscape app. So we have the pathway enrichment map. And we have a pathway of interest here. It's non coding RNA metabolic process. And then we can use gene mania to create a gene gene network. So here a node is a pathway and gene mania a node is a gene. And then we do that to get more detailed information about a pathway of interest. And here we can see that using gene mania and the function, function enrichment of gene mania, we could detailed the pathway into one, two, three, four subgroups of more defined pathways than this most significant node, the non coding RNA metabolic process could be divided in ribosomal RNA metabolic process, tRNA, DNA template transcription initiation large ribosome units. And we could color the nodes that are now gene based on the relative to gene expression. So as a final slides, so gene mania can be, so works with a few model organism. If you would work with a non-modal organism and you would have to create this gene function prediction network, then it's possible to do. You can constrict it with your model organism and gene mania can be used for other uses. So we use it for protein protein interaction network, but you could take the algorithm and modify it. And it's a work that was done in the lab of Gary Badder by Shroda Paye, and I think Ruth was working on it as well. So NetDEX is a patient classifier. So the networks here are different. And so that can be used to classify patients as well. So it's another option that you can do, but it's advent stopping because you have to recreate. You take the algorithm, but you have to recreate the database that are behind it.