Network Biology: Introduction to STRING and Cytoscape

[@science_mbg]

Thank you very much for the video and video starts at 24:30

[@shbangash437]

Incredible. This video clear all my concept . I really appreciate 🙏 your effort sir . Thanks for this session on KZbin .

[@larsjuhljensen]

Happy to hear that! :-)

[@Biomeducated]

24:45 START + INTRO + TOPICS in this webinar 33:47 STRING DATABASE and its features 47:51 How are various databases integrated in STRING? (Physical interactions, Co-expression data, etc.) 59:20 Break 1 (5 min) - Q's answered from chat box 1:04:22 How are various databases integrated in STRING continued. - Specific zoom in on 'Text Mining' from LITERATURE! 1:12:45 Knowledge Graphs - How all this mapped data is visualized 1:16:19 Answering questions 1:41:20 Overview of the web-based EXERCISES + Questions 2:01:55 EXERCISE 1 explanation 2:10:26 CYTOSCAPE visualization of STRING data 2:22:18 Use clustering (ClusterMaker2 app) for better interpretation of the network 2:22:58 SUMMARY + Acknowledgements 2:28:23 Example to interrogate PubMed for 'antibiotic resistance in S. pyogenes' in CytoScape + Network customization 2:35:25 Tutorial with Spreadsheet to Network 2:41:10 How to combine 2 networks? 2:46:40 Use of ClusterMaker2 2:50:12 Quest for questions 2:55:30 Mapping of compounds to proteins 3:08:17 Close-off *Thank you so so much for sharing this, Prof. Jensen!* Sharing stuff like this will accelerate scientific (and drug) discovery. I'm going to dive into my RNA-seq DEG lists with this new knowledge RIGHT NOW! So excited!

[@larsjuhljensen]

Thank you!

[@rufel9332]

Thank you so muchhh

[@larsjuhljensen]

You're very welcome, I hope it was useful!

[@jairsoto6150]

Chingón, carnal. Gracias por compartir tu experiencia, me ha sido de mucha ayuda.

[@larsjuhljensen]

Glad it was helpful :-)

[@Saed7630]

Very informative talk/presentation!

[@robertofernandorosalesroja5380]

Great demo, awesome talk!!!! Thanks!

[@larsjuhljensen]

Thank you very much - glad you liked it!

[@TheManiaKyu]

demo work really helped a lot. thank you very much~

[@supatkhongfak8022]

Thank you very very much

[@larsjuhljensen]

You are most welcome!

[@antonyshadowbanned]

Brilliant presentations, thank you so much

[@larsjuhljensen]

Glad you like them!

[@nidhidubey4798]

sir, if i have a set of genes with me. Can I make a gene regulatory network using cytoscape?

[@larsjuhljensen]

Yes and no. Cytoscape is a tool for analysis of networks and visualization of them, so Cytoscape itself is not going to make any any kind of network from a set of genes. However, if you have a source of regulatory interactions for your genes, you can easily import those into Cytoscape and visualize it (and it is a very good tool for that). Also, we are working on adding regulatory interactions to STRING, so in the future you will be able to import obtain a regulatory network in Cytoscape by using stringApp.

[@masoudrezaei331]

this video was awesome! wish you the best professor!

[@larsjuhljensen]

Thanks!

[@KainBAl]

Thanks Sir. This video has been very educational for me, I am very new to using these tools. I have a question, once I have my network in cytoscape and I have different groups of proteins separated into clusters, how can I assign a specific biological function to each of these clusters.

[@larsjuhljensen]

What you can do is to select all proteins in one cluster and then run the stringApp enrichment analysis on the selection only, using the full network as the statistical background. That will give you enriched terms for the selected cluster. It's bit of a faff, but you can obviously repeat this for all of the clusters, one at a time. I wish there was a way to automatically do it for all clusters, but that is not presently possible.

[@meenalbhardwaj4883]

is there any upcoming webinar for Cytoscape, string, ClueGo, MCODE I want to learn this tool for my master's project. the session should include live interaction if possible

[@larsjuhljensen]

I have already recorded a brief introduction to Cytoscape, which I plan to publish either Friday this week or Monday next week. I do not plan to cover ClueGo specifically, since I do not use it myself, but I will be covering enrichment analysis. Also, I intend to make a presentation about network visualization, which should include both network layout and network clustering (although likely not MCODE specifically). For a practical part, I recommend that you take a look at the online exercises at jensenlab.org/training/stringapp/

[@sachinsahu533]

How to prepare the Protein-Protein Interaction network, and which software is used to prepare for the network.

[@larsjuhljensen]

I'm not sure what network you want to prepare. As described in the presentation, you can get a network for a set of proteins from STRING. Not sure what you mean by "prepare", since these networks already exist. But if you want to analyze the networks, Cytoscape would be the tool of choice.

[@sachinsahu533]

@@larsjuhljensen Thank you Sir

[@mamoketebokhale6283]

Hi Prof Jensen, I had an input of 30 proteins (upregulated genes) and got only 3 interactions ( out of 15 items of the output) on STRING, in this case do i expand and add more interactors on the shell? Is it also advisable to hide disconnected nodes

[@mamoketebokhale6283]

That is when i used high confidence (0.07)

[@larsjuhljensen]

If you started with 30 proteins and got only 15 in your network to start with, it sounds like you have had some issues with the names you used for searching. That would be the first thing that I would check. Regarding number of interactions, which organism are you working on? The density of the network varies greatly between organisms; in some 0.7 may be a quite high confidence cutoff to use.

[@mamoketebokhale6283]

@@larsjuhljensen Thank you so much Prof, i mentioned on one of your videos that I work with Pectobacterium. You are very correct it seems for some of the names it didnt recognise them. I would do further searches on the names.

[@shahabmirshahvaladi6239]

great tak

[@shahabmirshahvaladi6239]

i wish the string viruses database included COVID proteins as well, comes real handy these days!

[@lichuinchong6161]

For STRING network analysis, what is the difference between two clustering algos, k-mean and MCL? bcoz both are giving me different results, which one should I follow and/or better?

[@larsjuhljensen]

In my experience, MCL is the better algorithm for clustering networks. From a user perspective, the big difference is that you explicitly say how many clusters there should be when using k-means. Unless you choose that number wisely, k-means will give poor results.

[@vrundakumbhar5651]

Sir can you please leave the exercise link here on KZbin? Please please please 🙏

[@larsjuhljensen]

All exercises can be found at jensenlab.org/training/

[@vrundakumbhar5651]

Sir you have said a proteomic experiment and data is arrived from that... So I want to ask if we want some proteomic data can we get it direct without any experiment directly from proteomic databases?

[@larsjuhljensen]

Most often tables like this are available as supplementary information files in proteomics publications.

[@elisabettaversace932]

brilliant, thanks! As a novice, I would have 2 questions 1) when are high/medium/low interactions scores recommended? maybe there is a publication that explains this and I did not find it. If I have just 50 genes that come from the same experiment, shall I use medium or low of high? And I guess I should use the same criterion for different conditions/experiments in the same work, right? 2) shall I prioritise evidence from experiments in the interaction sources? I thought that maybe data mining can inflate some associations reporting what is already known multiple times, but maybe this is accounted for. What is the gold standard? Maybe there is an obvious paper as a reference for this... Thanks a lot

[@larsjuhljensen]

There is no simple answer to which confidence cutoff to use - it depends on a lot of things, including what you're trying to accomplish, how many genes you're looking at, and how well studied they are. For visualization purposes, it will always be a pragmatic choice: if the interactions that are actually interesting to you are all high confidence, you're probably best off not showing a lot of lower confidence interactions that just muddy the message. And yes, I would certainly not want to mix and match different cutoffs within one analysis. If, on the other hand, you are using the network as input for an analysis like clustering, I would almost always want to include all interactions even with the lowest confidence, but of course take the confidence scores appropriately into account in the analysis. The reason is that throwing away part of your data before even starting is rarely a wise choice.

[@larsjuhljensen]

If what you want is a functional association network, then the weighting of evidence is already inherently done. Since all evidence sources were benchmarked on pathway knowledge (from the KEGG database), the confidence scores already weight them correctly. There are cases when you would want to exclude some types of evidence, usually to avoid circular reasoning in downstream analysis. But I cannot think of a situation when I would want to use all of them but prioritize them differently. Not unless what you want is in fact a physical interaction network, but we now give you the option to have that instead of functional associations.

[@gayathrisudarsan9044]

I have one question regarding STRING. I have around 400 targets for which I need to collect only the pathways. When I try to input using the multiple proteins options, it forms a network of 256 proteins and I get around 137 pathways. What I need is the pathways for each target from the list of 400 and odd targets and it's fine if there are overlaps as I wish to construct a network using that. Is there a way of doing it?

[@larsjuhljensen]

If you go to the "Save / Export" section at the bottom of the "Analysis" page, you can save the enrichment results. It contains a column with the matching proteins for each term/pathway, thus allowing you to get the information you are requesting.

[@gayathrisudarsan9044]

@@larsjuhljensen I tried that and I got the results for all the 427 genes I had. Thank you!

[@chigurupalliraja1909]

hi sir, i am searching for protein-ligand interactions for cancer in cytoscape which is my project.. ex: protein is 1TUP and ligands are some lets say 10. i should use this cytoscape software and find best ITUP - 10 ligands. so if ligand1 is better for ITUP protein so the answer is ligand1. please help me regarding this

[@chigurupalliraja1909]

and my main work is finding targets for proteins so....importing proteins in one column and ligands in second column and importing them in cytoscape is fine but i am having doubt regarding how should i write those protien and ligand names in that excel sheet so that cytoscape can understand for example i have protein 1TUP and ligand as alpha pinene how should i write these in those columns sir please help me regarding this thank you sir

[@larsjuhljensen]

Here's the thing: Cytoscape doesn't know anything about proteins or ligands. So the names that you use only matter if you need to merge in other data, in which case you want to coordinate the names across whichever data sources you need to combine.

[@johirislam8174]

Hi i analyzed my PPI netwrok by clustalviz which is cytoscape plugin.Here the results came by rank.So now i am confused which one i should pick

[@larsjuhljensen]

Sorry, I don't understand your question. I'm not familiar with clustalviz.

[@ahmedhashim5783]

When I upload a table from File to visualize the netowrk, the columns appear blank and empty of content. I tried several file,, txt, xml, csv... and installed previous versions of Cytoscape, but the problem was not resolved!

[@larsjuhljensen]

I'm almost certain that your problem is that the column you selected as "key column" in the file you import does not match what is in the column you selected as "Key Column for Network". If the two do not match, none of the nodes in your network will match a row in the table that you're importing, and the result will be precisely what you describe: the new columns are created but they are all empty.

[@ahmedhashim5783]

@@larsjuhljensen The list of genes that I entered into String protein query is 5100, 4600 genes passed filter and the network was formed. The same excel sheet from which I took the gene list I used it as a table to visualize the network, (the goal is to visualize the upregulated and downregulated genes), the excel sheet contains in addition to the gene id column, some columns like fold change, p.value.. etc. When I emport the table, the key column is gene id with the shered name column in the network. The result; columns display only their titles but are empty

[@larsjuhljensen]

@@ahmedhashim5783 If you used stringApp to retrieve the STRING network, the gene ids from your Excel sheet will not be in the "shared name" column of the network. That column will instead contain the STRING protein ids. The terms you queried with (i.e. gene ids in your case) will be in the column called "query term". You can find a detailed explanation of this under jensenlab.org/training/stringapp/#33-data-import

[@ahmedhashim5783]

@@larsjuhljensen Exactly, I had do this yesterday and was succeeded. Thank you

[@zhanyouxu8325]

Thank you very much for the video. I have one question as below: When I tried to import network from public databases, under "data source:" I can see ONLY one option: universal interaction database client. I can NOT see other data sources. How can I add/see other data sources? Thank you.

[@larsjuhljensen]

You need to install stringApp from the Cytoscape app store to make the other options appear.

[@zhanyouxu8325]

I figured it out that I need to install the stringapp. thanks.

[@shashishekharsingh2685]

which file should i download from string to make PPI in cytoscape?? Please reply

[@larsjuhljensen]

No need to download any file. Just install stringApp from the Cytoscape app store. See exercises at jensenlab.org/training/stringapp/

[@stephanyroserorojas894]

HOW DO I INCREASE THE NAMES OF THE KEGG ROADS TO MY INTERACTION NETWORK? THANK YOU

[@larsjuhljensen]

Not sure what you mean, sorry.

[@shadyamr9775]

Hello, thanks a lot for the introduction. I have a question I tried to import the "antibiotic resistance streptococcus " as you did but it give an error and doesn't work?!

[@larsjuhljensen]

I cannot guess why that would be. I just tested and it still works fine for me.

[@shashishekharsingh2685]

How do i link proteome network to transcriptome data? if you can make video on that that will be very helpful.

[@larsjuhljensen]

If you have a list of differentially expressed genes that you want to map to a protein network, you'd do it exactly the same way as for a proteomics dataset: jensenlab.org/training/stringapp/#exercise-3

[@بلالجاسر-ج9ذ]

Hello dear Sir Thank you very much for the nice explanation Sir, may I get your help to build up my network then choose the most active protein for a specific disease?

[@larsjuhljensen]

It is not really clear to me what you mean by "most active" in this context.

[@بلالجاسر-ج9ذ]

How can I recognize the proteins to do docking from the results of cytoscape or string??

[@biogerontology7646]

55:00

[@bidwansekhar]

Take me as a PhD then

[@larsjuhljensen]

Sorry, I don't currently have open PhD position. But I advertise all positions at jensenlab.org/, and they are filled in open competition. I often have one or more strong candidates that I know will apply, but I always tell them that there is no guarantee they will get it. So feel free to apply when I advertise a position :-)