[SOUND] [BLANK_AUDIO] In the second part of this lecture I'd like to discuss, rather briefly two papers where. Taking large stroop of data and casting them as networks has allowed us to get an understanding of the system and answer this, and answer questions. The first paper deals with analysis of networks from differentially expressed genes. And what these researchers did was to study the effect of interferon which is protein called asyricrine that regulates immune cells and an antiviral. Agents. On the gene expression patterns and blood [INAUDIBLE]. Interferon is actually used as a drug to treat or interferon beta is used as a drug to treat multiple sclerosis, and in this study, the researchers Studied individuals treat a given interferon as a drug varying times after the drug had been administered. They had a So they studied two of these individual. And they had four control patients who did not have disease and were not given the drug. They drew blood from all these individuals. Made, made messenger RNA and then made CDNA. And then Did a microwave experiment and overall they found that they could identify some 438 genes that are differentially expressed in the patients but not in the control subjects, and this pattern is shown here. Red of course means that the genes went up. Green means that the genes went down. It was decreased from the control. Black means no change. You can see that these various changes of the genes can be plotted as a function of time and then clustered here, and I won't go over this clustering in great detail. To, to understand different kinds of relationships. So, the gene clusters. Once these genes are clustered, they can be analyzed for go or [UNKNOWN] terms. And the analysis of the patterns. Of the up and down regulated genes, suggested that cell depth is triggered by. In this case by regular, by down regulation or decrees in expression of anti-salidate genes. An interesting sort of mechanistic operation. So these others cast these differentially expressed genes as networks, using coordinate or Expression of these genes to represent edges. We used a concept called Mutual Information theory where information about one gene tells you something about the other gene. And this can be used to calculate the probability of co-expression of two genes and this The color in this particular network, the color of the edges such as gray, represents the strength of the correlation. Red is the strongest correlation, followed by yellow, followed by green, and by gray. And so one can look at how these different kinds of genes are expressed with respect to each other. There are two networks in this figure. One a network from an interferon beta treated patient. Which is A, and a network from control. Subjects which have not been treated by drugs. You can clearly see. Just visually that there are many more genes that are expressed in a correlated fashion in the interferon beta-treated pa, cells from the interferon beta-trea, treated individuals as compared to control individuals. These different genes could be color-coded with respect to their. A good categories and putting them in this kind of a framework here allows one to sort of see what kind of relationships there are between the different groups of genes that may be classified using [UNKNOWN] terms or how different sub cellular processes are co-regulated or regulated in a time dependent manner according to [UNKNOWN] beta treatment. Another paper in this case Using protein protein interaction used a similar sort of computational approach of building networks to understand sort of a biological question. In this in this study the researchers were interested in understanding how >> Infection with the hepatitis C virus are coordinated [UNKNOWN] within the cell, to do this what we studied was to look at the proteins from the hepatitis C virus and study its interactions with. The human proteins and one can see there can be different kinds of interactions the Hepatitis C virus. The virus protein can interact with human proteins that is the Edges we need the black and the red. And of course not all human proteins will interact with the widest proteins so there are going to be two classes of them, the ones go in the, shown as the red nodes that interact with the wild proteins, and those shown in the blue nodes which interact with the red nodes, but do not directly interact with the wild protein. So when they found that 11 viral proteins interacted with 421 human proteins and these human proteins could be classified in terms of gene ontology terms and then used to double up an understanding of what sorts of sub-cellular processes and or genes are involved in recognition. Reaction to the hepatitis C virus and such. So, when these people analyzed these 421 G proteins and classified them with respect to sort of The gene ontology term that they found that the physiological data suggested in the multiple signaling pathways here: insulin cytokine, the so called JAK/STAT pathway, and the transforming growth factor, all of which interact with the human. It's [UNKNOWN] so different kinds of signaling pathways appear to have interactions for with a human hepatitis virus. So the question could, well one could as the question this is the case do the human proteins that interact with these HCV proteins act as connectors between these pathways and the network building that is done here. Suggests that indeed this is the case. You can see that from the JAK/STAT pathway to the insulin pathway and so on, that a lot of the signals can go through the intermediate [INAUDIBLE] which are the proteins that interact with the HCV, proteins. So this [UNKNOWN]. So, one rule or one method by which, hepatitis C virus can sort of hijack cellular function, is to interact with. Proteins that are connectors between multiple different signalling pathways. So when they took this kind of analysis and looked for new pathways or new sub-networks that might new subnetworks that might be involved in >> Interaction between the viral proteins between the virus and it's target cells they form a cell network which [UNKNOWN] focal adhesions which are points of contact, focal adhesions which are point of contact with the extracellular matrix. So these authors ask the question do the viral proteins affect the ability of the cell to bind to the extra extracellular matrix? So using the experimental approaches these authors were able to show that the answer is actually yes. And this kind of, of determining how. Viral proteins may hijack cell, cell interactions or cell matrix interaction, promote infectivity can be understood by developing these kinds of networks from protein, protein interaction data. It should be said that the pathway annotation and network building can thus be used as a Tool to discover the involvement of a new subcellular process in physiological or pathophysiological function. In this case, it's the [UNKNOWN] molecules in terms of [UNKNOWN] we interact. [INAUDIBLE]. So combining the annotation of nodes with network topology that is using good terms or [INAUDIBLE] terms. There is [INAUDIBLE] network topology allows us to build networks that Can be, very useful in identifying sub-cellular processes involved in a cell or tissue function. Databases such as Gene Ontology or KEGG are indeed needed to annotate these gene functions, and so you can readily see how database And bioinformatics usage is very important, in conjunction with network analysis to understand, to obtain a functional or mechanistic understanding. From these two examples, you can also see that building undirected network based on different criteria includes co-expression of genes in network one. A protein-protein interactions allow us to understand how different sub-cellular processes can interact. In the context of physiology or pathophysiology. So, the take home points for lecture nine are the following: network properties are useful in understanding how large systems are organized, and both the global and the local properties of networks are useful in understanding the organization of Large systems, such as another network is scale free, suggesting that it would be robust but that it does, has a high glass strength. A coefficient suggesting that it would be groups or modules of interacting components that are functional within the network. From lists of nodes, cell biological networks can be built with different criteria. So I showed you one example where the edges are presented co-expression of genes, and another example where the edges are presented protein protein interactions. So it's not only necessary that one has only one type of edges but can have multiple types of edges, edges in building networks. Network building usually involves combining data from high throughput experiments and interaction with prior knowledge. You can see that work in this case of differential express genes, using the differential express genes and using the go terms itself. Combines, data from hyper experiment and the good terms that present sort of prior knowledge. So, you need both of these to sort of build and analyze networks. And these networks have a real interrelationship between different signalling pathways and different sub-cellular process, such that one can see that these multiple, many sub-cellular processes can come together to pro, to, to sort of yield, define phenotype. So, indeed interrelationships between modules, between sub-cellular functions is a key. Sort of insight undirected networks can provide us. Thank you. [MUSIC]