Lesson todos

[nsohail19]

All lessons

• Zenodo citation missing

Lesson 01 - Spatial Transcriptomics Technologies

• (optional) Maybe a few exercises of scenarios where people would pick a technology to use for the research question

Lesson 05 - QC

• Maybe move the Spatial overlay to the second tab since we don't really discuss it and the discussion above is about the plot we don't see by default? NS note: I like having the visual upfront but I understand why the suggestion is to put it second. I am going to leave it as is and we can revisit based on feedback from the class

Lesson 10 - Seurat Cheatsheet

• "The following requested variables were not found: Spatial.008um_PTPRC" when using FetchData NS note: did not get this error
• On DimPlot do we like just the first few clusters being annotated?

Lesson 15 - Cellchat

• "In essence, the method identifies overexpressed ligand receptors from the database in your own data to measure association using the principle of mass action." Is this the same as the Law of Mass Action? Also, maybe explain this principle.
• For the callout on how to subset the the CellChatDB, I think it might be better to show the example of subsetting to two categories but then providing the terms needed to subset from each of the four categories ("Secreted Signaling","ECM-Receptor","Cell-Cell Contact","Non-protein Signaling"). It also isn't clear to me the difference between non_protein argument is controlling when "Non-protein Signaling" exists as a categories to search for?
• "These scores are calculated with a wilcoxon and auROC test from the calculateOverExpressed() function." Do you mean the identifyOverExpressed()?
• "This is very similar to how we ran FindMarkers() in a previous lesson. The results are stored in the @var.features slot of the CellChat object." I think maybe the first sentence should get moved up to the previous paragraph because the way it is written makes it sound like the results being stored in var.features is how CellChat is similar to FindMarkers, but I think you mean the methods used to measure expression are similar to FindMarkers?
• My Table 3 results are almost entirely different from yours
• "Summarize gene expression per cell type by average and normalize (triMean)" What do we mean by triMean?
• I feel like the steps in the computeCommunProb() workflow are not super clear and, for me, it is hard to follow. The steps don't feel like they connect super well.
• Can we provide more of a description of what scale.distance does for computeCommunProb()
• "Those three steps take a long time to run, so we have created an RDS object for you to load in with the results from these steps:" Can we give a range on time estimates?
• "We can now filter results based first on the how many bins belong to each cell type:" I don't think I understand this step clearly because we would know the bins that were below that threshold before running CellChat, so I don't understand why we'd run it after and I can't imagine any of our called cell types didn't mean the 5 cell threshold that we set in the command NS note: this is to fool-proof the method so that if someone else ran it and they did not have enough cells, it would automatically be removed.
• "Each of the ligands and receptors are associated with different signaling pathways.", should this perhaps be "Many of the ligands and receptors are associated with multiple signaling pathways." NS: this is not true, each LR pair is mapped to a single pathway
• "Each of the ligands and receptors are associated with different signaling pathways. So to summarize the results, we can aggregate the values to create a pathway-level score to identify more general trends in the data." This isn't clear to me. It would feel like if we have a pathway level score that we are aggregating, then that would be because multiple ligands and receptors are in the SAME pathway, not different pathways? NS: this is not true, each LR pair is mapped to a single pathway
• "We can also see that fibroblasts appear to be sending signals to tumor cells." Is this any more noteworthy thanIntestinal Epithelial cells interactions with tumor cells or is the story just better? The heatmap is more convincing on this argument IMO NS: But the heatmap has the 2nd highest score for fibroblasts and tumor cells?
• It's fine and good to have, but with our dataset the gg_h_weight plot is sort of boring. Feels like the Tumor cells are maybe drowning out signal between other sources and receivers. I don't think we can fix this without subsetting or something, but mostly just commentary
• "The last steps we ran in the CellChat workflow was to calculate statistics for the overarching pathways involved with each ligand-receptor." As before, I am not getting this. Aren't these pathways made up of multiple signaling systems?
• "So now we can take a look at the ranked, top significant pathways" something about "ranked, top significant" sounds off to me
• Can we get a y-axis label for Figure 6 and a scale for the x-axis? NS: cannot be done easily
• My Figure 7 had no labels in RStudio
• Explain what the netVisual_aggregate() shows
• netVisual_aggregate plot gave me this error "Error: not enough space for cells at track index '1'."
• Thoughts on moving the creating of n_ct from the top_pathways codeblock to fig-netVisiual_bubble code block since we don't use it in the top_pathways codeblock?
• "This is represented as bubbleplot, where the size of the dot is proportion to the p-value." An adjusted p-value correct?
• All of the dots in the bubble plot are the same size?
• "So we will look at all ligand-receptor complexes in those top pathways that have tumor cells as the source." and "Bubbleplot of significant pathways with tumor cells as source" The direction of the arrows on the x-axis make it look like tumor is the target not the source?
• I cant render the lesson because I don't have "15_cellchat_crc_computeCommunProb.qs"
• For the sessionInfo, what are you thoughts on loading all of out packages rather than just the packages from the current R session? Feels like it would be more in line with reproducibility? I sort of feel like Ideally it would happen at the end of each Rscript, but we don't have that. Adjust as you see fit.
• I have added some verbiage around the sessionInfo section, feel free to edit or add to it as you like