Great videos! Thank you for sharing them with us! I have a quick question - When we provide as input multiple proteins, the first image we get is all the input proteins connected (or not) by either evidence or confidence. My question is - the localization of each protein in this image has some kind of meaning or is it random? I was wondering if perhaps it would mean that some proteins may be closer to each other because they share some specific features even if they are not connected by evidence/confidence. Thank you!
You're welcome! The short answer to your question is that the position does not mean anything. It is not random - it is generated by a layout algorithm. I have a short video on the topic: czcams.com/users/shortsF3Nrqjxlw4A?feature=share
Thank you for your great instruction video. I wonder how to interpret the data from the enrichment analysis? For instance, say i've entered a list of differentially expressed proteins from a proteomics study. How would you select relevant enriched processes, on 'strength', or is count important also? How does one deal with overlapping GO terms or pathways? What would you advise on the FDR cut-off (
Great question - I wish there was a simple answer. Finding the relevant terms really comes down to knowing the biological question that you are working on; I do not believe there is an automatic approach to say which terms are more relevant. But when it comes to the redundancy stemming from overlapping GO terms and pathways, STRING allows you to see that. You can click on multiple enriched terms of interest and get the genes highlighted in the network. That allows you to easily see if multiple terms in fact stem from the same set of genes. Regarding FDR, there is again no right value, but you will generally want to focus on the most enriched terms on and not the ones far down the list.
@@larsjuhljensen I have a follow-up question and I would be much obliged to hear your opinion. Say i have a large set of tested proteins, but not genome-wide, and a fold change: should one apply the 'multiple proteins' function with a set of significantly regulated proteins or is the better alternative to apply the 'proteins with values/ranks' functionality with all tested proteins and FC?
Not directly, since STRING is a database of protein-protein interactions and thus has no knowledge of metabolites. A solution to could be to use Cytoscape stringApp instead, as it integrates the latest STRING network with data on protein-chemical interactions from the STITCH database (which are unfortunately not very up-to-date, but should be fine for well-known metabolites).
@@DrAkhtarAli08 I've never tried it, since I don't work on metabolomics. As far as I could tell from a very quick search, it wouldn't be linking metabolites to proteins but rather metabolites to each other. But I could be wrong.
Thank you sir , you made all our lives easier!!!
Happy hear that - thank you! :-)
Great videos! Thank you for sharing them with us! I have a quick question - When we provide as input multiple proteins, the first image we get is all the input proteins connected (or not) by either evidence or confidence. My question is - the localization of each protein in this image has some kind of meaning or is it random? I was wondering if perhaps it would mean that some proteins may be closer to each other because they share some specific features even if they are not connected by evidence/confidence. Thank you!
You're welcome! The short answer to your question is that the position does not mean anything. It is not random - it is generated by a layout algorithm. I have a short video on the topic: czcams.com/users/shortsF3Nrqjxlw4A?feature=share
Thank you for sharing your knowledge, hope you are doing well sir!
Thank you, Dr. Jensen!
You are most welcome :-)
Thank you very much for these amazing presentations to understand the use of STRING
You are very welcome - as a STRING developer, I really want people to know how to use the tools we make the best way.
wonderful explanation, thank you so much Professor Jensen!
You are very welcome!
@@larsjuhljensen looking forward to your future videos!
Awesome smart stuff
Thanks 😊 sir....exclusive work
Wao...great...
Thank you for your great instruction video. I wonder how to interpret the data from the enrichment analysis?
For instance, say i've entered a list of differentially expressed proteins from a proteomics study. How would you select relevant enriched processes, on 'strength', or is count important also? How does one deal with overlapping GO terms or pathways? What would you advise on the FDR cut-off (
Great question - I wish there was a simple answer. Finding the relevant terms really comes down to knowing the biological question that you are working on; I do not believe there is an automatic approach to say which terms are more relevant. But when it comes to the redundancy stemming from overlapping GO terms and pathways, STRING allows you to see that. You can click on multiple enriched terms of interest and get the genes highlighted in the network. That allows you to easily see if multiple terms in fact stem from the same set of genes. Regarding FDR, there is again no right value, but you will generally want to focus on the most enriched terms on and not the ones far down the list.
@@larsjuhljensen Thanks very much for your response!
@@larsjuhljensen I have a follow-up question and I would be much obliged to hear your opinion. Say i have a large set of tested proteins, but not genome-wide, and a fold change: should one apply the 'multiple proteins' function with a set of significantly regulated proteins or is the better alternative to apply the 'proteins with values/ranks' functionality with all tested proteins and FC?
How if i want to find an interactions of view protein only, can you show us how?
Sorry, I don't understand your question.
Can we use "string" for metabolomics?
Not directly, since STRING is a database of protein-protein interactions and thus has no knowledge of metabolites. A solution to could be to use Cytoscape stringApp instead, as it integrates the latest STRING network with data on protein-chemical interactions from the STITCH database (which are unfortunately not very up-to-date, but should be fine for well-known metabolites).
@@larsjuhljensen what about GNPS networking? Have you used?
@@larsjuhljensen Thanks for letting me know
@@DrAkhtarAli08 I've never tried it, since I don't work on metabolomics. As far as I could tell from a very quick search, it wouldn't be linking metabolites to proteins but rather metabolites to each other. But I could be wrong.
@@larsjuhljensen ok thanks
your eyes r pretty😍