You briefly discussed mapping chromosome number to the node/vertix, what significance/use-case do you see for this? Do you know of any transcription factors (TFs) that regulate target genes of a specific chromosome more than another? I know that certain developmental genes follow a developmental pattern like the HOX genes. For example, HoxA genes are located on chr 6 and HoxD on chr 2. Could a GRN visualization help someone see these types of a patterns?
Secondly, you mentioned centrality in technical terms but is there any biological relevance to using this statistic? There’s a debate (particularly for protein-protein networks) that centrality represents ‘essential’ genes, is there anything else you know of?
(1) In terms of chromosome number pathways I am not immediately aware of specific transcription factors associated with regulating a chromosome more. But if those patterns exist, it could definitely be explored with a GRN. The use case I was actually thinking about is actually inspired by the Keurentjes et al. paper we curated that uses eQTL data (combined with expression trait profiling and gene annotation) to elucidate the gene regulatory network. Chromosome location is a pretty rough measure but it might suggest eQLT which might suggest co-regulation, or at least warrant further investigation into location.
(2) For centrality in general across different species betweenness centrality is negatively correlated with non-synonymous site mutation (Kim et al., 2007) so a form of “essentialness”. Another paper Li et al., 2017 found that effectors manipulated networks at positions with higher betweenness centrality. But these are all just other ways of saying “essential”. I know other studies have used GO-terms combined with betweenness centrality to identify key points of regulation, but GO-terms less essential for aGENT since the data is already curated by network type (Liseron-Monfils et al., 2015).
(3) Next steps includes feature pruning (to prevent featuritis as mentioned in your thesis), user testing, and creation of demonstration content and tooltips to explain graph theory terms that are more obscure in the field.
An important node is a subjective term. It could mean a node that the user is interested in, be tagged as important in curation, or network statistics and/or expression levels may suggest that it is important in that network. There is no cut off per se. Thanks for your question!
Hey Rachel. Really impressed with what you’ve built! Like Vincent, I’m also curious about the future directions for AGENT. Best of luck in grad school! (:
Thanks for your kind words! For future directions we were supposed to be doing user testing but that was interrupted by corona virus. We are still planning on doing more informal user testing and refining the features we do have. Have a great summer!
Great presentation! For the Selected Targets functionality, the chart seems to mostly have interactions in a diagonal from the top left to the bottom right – is this just that particular example, or is it a pattern that always shows up? I ask because if it’s a patter, it seems like the chart could be reduced to a correspondence list or something like that instead of having so many empty cells (although maybe the format is for compatibility with other tools, I’m not sure)
Great question! The diagonal is just for that example. But yes, the chart could be reduced to a list. This visualization is based on something previously in AVI2 but there is no particular reason it has to be a chart and not a list. I do think that an interactive sortable (where you can filter and cluster by different fields like source, target, interaction type ect.) sif file would be a good (or even better) substitute since the number of items selected is up to the user.
Hey Rachel! Your project seems great, I really appreciated the background you gave and detailed walkthrough of how the tool works. I was wondering what shortest path algorithm you chose for making the network and how you chose it?
Hi Sakshi, I used Dijkstra’s algorithm as it was built in to the CytoscapeJS library. It is also the standard in terms of fastest for shortest path when you don’t know how dense the edges are.
Good job, Rach!
You briefly discussed mapping chromosome number to the node/vertix, what significance/use-case do you see for this? Do you know of any transcription factors (TFs) that regulate target genes of a specific chromosome more than another? I know that certain developmental genes follow a developmental pattern like the HOX genes. For example, HoxA genes are located on chr 6 and HoxD on chr 2. Could a GRN visualization help someone see these types of a patterns?
Secondly, you mentioned centrality in technical terms but is there any biological relevance to using this statistic? There’s a debate (particularly for protein-protein networks) that centrality represents ‘essential’ genes, is there anything else you know of?
Lastly, any future directions for AGENT?
Hi Vincent!
Thanks for your questions!
(1) In terms of chromosome number pathways I am not immediately aware of specific transcription factors associated with regulating a chromosome more. But if those patterns exist, it could definitely be explored with a GRN. The use case I was actually thinking about is actually inspired by the Keurentjes et al. paper we curated that uses eQTL data (combined with expression trait profiling and gene annotation) to elucidate the gene regulatory network. Chromosome location is a pretty rough measure but it might suggest eQLT which might suggest co-regulation, or at least warrant further investigation into location.
(2) For centrality in general across different species betweenness centrality is negatively correlated with non-synonymous site mutation (Kim et al., 2007) so a form of “essentialness”. Another paper Li et al., 2017 found that effectors manipulated networks at positions with higher betweenness centrality. But these are all just other ways of saying “essential”. I know other studies have used GO-terms combined with betweenness centrality to identify key points of regulation, but GO-terms less essential for aGENT since the data is already curated by network type (Liseron-Monfils et al., 2015).
(3) Next steps includes feature pruning (to prevent featuritis as mentioned in your thesis), user testing, and creation of demonstration content and tooltips to explain graph theory terms that are more obscure in the field.
Thanks for a great year and all your help!
Rachel
nice explanation. What is the cut off for important node ? Does it involve statistic calcluations of the data?
Hi Wen Kai,
An important node is a subjective term. It could mean a node that the user is interested in, be tagged as important in curation, or network statistics and/or expression levels may suggest that it is important in that network. There is no cut off per se. Thanks for your question!
Hey Rachel. Really impressed with what you’ve built! Like Vincent, I’m also curious about the future directions for AGENT. Best of luck in grad school! (:
Hi Ida,
Thanks for your kind words! For future directions we were supposed to be doing user testing but that was interrupted by corona virus. We are still planning on doing more informal user testing and refining the features we do have. Have a great summer!
Hi Rachel,
Great presentation! For the Selected Targets functionality, the chart seems to mostly have interactions in a diagonal from the top left to the bottom right – is this just that particular example, or is it a pattern that always shows up? I ask because if it’s a patter, it seems like the chart could be reduced to a correspondence list or something like that instead of having so many empty cells (although maybe the format is for compatibility with other tools, I’m not sure)
Hey Bruno,
Great question! The diagonal is just for that example. But yes, the chart could be reduced to a list. This visualization is based on something previously in AVI2 but there is no particular reason it has to be a chart and not a list. I do think that an interactive sortable (where you can filter and cluster by different fields like source, target, interaction type ect.) sif file would be a good (or even better) substitute since the number of items selected is up to the user.
Hey Rachel! Your project seems great, I really appreciated the background you gave and detailed walkthrough of how the tool works. I was wondering what shortest path algorithm you chose for making the network and how you chose it?
Hi Sakshi, I used Dijkstra’s algorithm as it was built in to the CytoscapeJS library. It is also the standard in terms of fastest for shortest path when you don’t know how dense the edges are.