Definition of 'Specific Pathways' and minimizing pathways/cell-cell interacting molecules in a plot

[Yanst86]

Hi,,

First of all, thanks for the great package, it provides very interesting biological insights about my data! I have two technical questions about plots that compare two datasets:

1. In "netAnalysis_signalingChanges_scatter" - how does the algorithm define that a pathway is specific to a certain dataset? For example, in the attached , what could be the reason that PECAM1 is considered specific to a certain dataset but not FN1 or MHC-II which have higher differential incoming and outgoing signaling scores?
2. I have a lot of cell-cell interactions and pathways that were identified in my analysis, and the plots that I try to present, e.g. rankNet, Signaling Role heatmap and especially the circos plots are unintelligible because the text boxes overlap. Is there any way to present only the top 5/10 pathways or interactions (according to p-value or another differential score) in these plots?

Thanks in advance, Yan

[sqjin]

@Yanst86 Positive values mean increased signaling in the second condition, and negative values mean increased signaling in the first condition. specific means it only appears in one condition for either outgoing or incoming.

You can check the tutorial where we show that differetial expression analysis can be combined.

[Yanst86]

Hi,

Thanks for the clarification for the first question. However, my 2nd question is not answered, I will try to clarify and give examples: My dataset consists of about 28k cells with 12 different cell groups. Because I have too many interactions and pathways emerging in my analysis, I have unintelligible plots, here are some examples:

Is there any way I can only show the top 5/10/20 Pathways or molecules according to p-value, communication probability or some other score so I have fewer pathways (therefore readable text) in each plot?

Thanks, Yan

[sqjin]

@Yanst86 For the second and the third ones, you can check the tutorial where we show that differetial expression analysis can be combined (Comparison analysis of multiple datasets using CellChat). For the last one, you may can only show the signaling with significant changed (the signaling marked as black are non-significant). You can extract all signaling gg <- rankNet(...); df <- gg$data and then specify the signaling = NULL, pairLR = NULL when running rankNet.

For the first one, I think you can only show highlight signaling that you are interest.