Add cell-cell communication chapter #140
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
|
Awesome, this is really cool! I'll start with some comments in a moment :)
Surprises me. They're usually somewhat instant for me. I'll keep an eye on it.
Yeah, that would be great. I fully trust you to pick someone that is competent but also enjoys reviewing such work. |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:25Z This section is usually always called "motivation" in our book. Please rename it although "background" is of course a perfectly valid name. |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:26Z transcritpomics
-> transcriptomics
commonly referred as ligand-receptor
-> commonly referred to as ligand-receptor
plagued by the lack of ground truth
-> plagued by a lack of ground truth
You can mention at the end why we're now showing which methods then. |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:28Z Please talk to Anna about the figure. She will redraw it. |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:29Z Rename to "Environment setup", please. |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:29Z Citation isn't done :) |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:30Z TODO: replace with pertpy dataloader soon. Likely next week or so. |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:31Z Line #1. # Check cell types Rather write some text above such as:
"We are examining the cell types for x y z reason". |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:32Z Line #1. # show UMAP as a refresher Rather write some text above as in: "Let's look at xyz for abc reason first" |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:33Z Explain these choices. |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:34Z Explain these choices |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:35Z Can you kill some of these warnings upstream? :) dbdimitrov commented on 2023-01-11T14:40:21Z Yep, updated liana to get rid of those :) |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:36Z Visual exploration
or something else. At least without the typos :) dbdimitrov commented on 2023-01-11T09:08:16Z hahahaha that typo is a work of art |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:37Z If you're not going to reuse the plot you don't need to assign it to a variable. |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:38Z specicity
specificity |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:39Z distrubtion
distribution |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:40Z See above. |
|
View / edit / reply to this conversation on ReviewNB Zethson commented on 2023-01-10T21:32:41Z I like this section. Well done. |
|
View / edit / reply to this conversation on ReviewNB grst commented on 2023-01-12T10:27:23Z Personally, I think comparisons between conditions are quite important and I'd even argue it's what most of the people want most of the time.
Therefore, I'd at least love to see that outlook section expanded a bit beyond a reference to the papers (e.g. summarizing what the methods do and linking to their tutorials).
Also, as far as I understood, NicheNet, as applied in this tutorial is also based on comparison between conditions, so maybe worth referencing this here. dbdimitrov commented on 2023-01-19T19:25:03Z Yep. Great point, I will adapt the NicheNet part to reflect that. dbdimitrov commented on 2023-01-22T14:31:33Z I also extended the outlook a bit in this direction, but honestly I wouldn't personally apply most methods. Many of them focus on interaction numbers between cell types (which I don't think is super meaningful) and few approaches were actually benchmarked. |
|
View / edit / reply to this conversation on ReviewNB grst commented on 2023-01-12T10:27:24Z
|
|
View / edit / reply to this conversation on ReviewNB grst commented on 2023-01-12T10:27:25Z hypothesis-free systems-level hypotheses
sounds a bit weird dbdimitrov commented on 2023-01-20T14:21:42Z "to generate systems-level insights in a hypothesis-free manner" should be clearer :) |
|
View / edit / reply to this conversation on ReviewNB grst commented on 2023-01-12T10:27:26Z NicheNet first employs a network algorithm, known as Personalized PageRank, to estimate the probability of a random walk from a ligand to terminate at a given regulator, or it estimates the signalling probability of a given ligand modulating a particular regulator.
I don't get how the "estimates the signalling probability of a given ligand modulating a particular regulator" connects to the random walk. Is this an alternative approach? Then how does it estimate the probability?
dbdimitrov commented on 2023-01-20T15:43:25Z yep, extended it a bit :) |
|
View / edit / reply to this conversation on ReviewNB grst commented on 2023-01-12T10:27:27Z
---
Comparison between conditions: is CellphoneDB v3 one of them? dbdimitrov commented on 2023-01-22T10:00:47Z I believe the v3 version of CPDB asks the user for DE genes and returns interactions which contain those. However, even if these are DEGs between conditions, I'm not sure if it is an actual method to do comparison between conditions. |
|
Really nice write-up, @dbdimitrov! |
|
Yep. Great point, I will adapt the NicheNet part to reflect that. View entire conversation on ReviewNB |
|
View / edit / reply to this conversation on ReviewNB dbdimitrov commented on 2023-01-19T19:25:18Z Prior to that should add technical considerations, expression proportions, complexes, stable cell types, etc. |
|
View / edit / reply to this conversation on ReviewNB dbdimitrov commented on 2023-01-19T19:25:19Z Robin: Other way around would be clearer, e.g. multiple senders, only one receiver. |
|
"to generate systems-level insights in a hypothesis-free manner" should be clearer :) View entire conversation on ReviewNB |
|
yep, extended it a bit :) View entire conversation on ReviewNB |
|
OK. Fair point, I can load e.g. dplyr and tdyr for my part alone, but then NicheNet includes most of tidyverse as imports, and Seurat (which itself imports half of CRAN and 90% of BioConductor).
View entire conversation on ReviewNB |
|
I believe the v3 version of CPDB asks the user for DE genes and returns interactions which contain those. However, even if these are DEGs between conditions, I'm not sure if it is an actual method to do comparison between conditions. View entire conversation on ReviewNB |
|
Updated these with Anna's figures :) View entire conversation on ReviewNB |
|
I also extended the outlook a bit in this direction, but honestly I wouldn't personally apply most methods. Many of them focus on interaction numbers between cell types (which I don't think is super meaningful) and few approaches were actually benchmarked. View entire conversation on ReviewNB |
|
Changed this and reduces some of the dependencies in the env
View entire conversation on ReviewNB |
|
@dbdimitrov, I'd definitely like to hear more about your thoughts on Cell2cell comparisons between conditions, but this is not the right place to discuss it! Maybe I'll open an issue in Lianapy? |
|
@grst yes, please do so :) We have been working on bridging the two (liana x Tensor) a lot lately, so I plan to add a tutorial soon. We're also working on a bit more in-depth tutorials, and I hope these will be available in the next month or so. |
Hi @AnnaChristina @Zethson,
As discussed, this is my draft for the CCC chapter.
In short, I give a general background on CCC, run CellPhoneDB, LIANA and NicheNet; summarize some of the assumptions and limitations of CCC in single-cell transcriptomics, and give an outlook with a few questions as a quiz.
I included a couple of example figures, I guess they don't exactly fit the style of the book. Nevertheless, the ideas are there, even though my artistic skills are not :)
Several things to note:
If we're looking for a second opinion, we could ask Erick Armingol, Hratch Baghdassarian, or Robin Browaeys.
Once I get some feedback, I'm happy to go over the chapter again to address any comments and fix potential typos, etc.