Consolidated feedback for improvements

Below is most feedback (re: improvements) over both the June BH and Aug CT courses. They are organized based on general areas / sections. They have been assigned to individuals. But, I will break these out and separately assign to each of you.

General (All – for your respective sections)
1. Few slides before each section to give a conceptual overview (Aug). Sue – add your figure to each section, highlighting where we are at.
2. A deeper higher-level understanding would be great for each step, i.e., differences between each normalization method. (Aug)
3. Similar comment specifically about deconvolution: I was a bit confused as to why we were doing each step of deconvolution. (Aug)
4. Another thing that you did well but could do even more of is summarizing before we switch to a new topic or instructor! To do noted by Dan/Sue/Brian: let’s revisit our objectives and take aways for each section. These could be more meaningful and informative.
5. A little more description about why we are doing each step, as I have very little omics experience. (Aug)
6. Participants would benefit more from the course if instructors provided more explanations of the science and justifications for adopting approaches or parameters where necessary (Aug).
7. Add comments to code to consolidate what we are going to do. (Aug) Dan did this, but others didn’t.
8. The pacing is a bit fast, mostly the explanation of the purpose of each function. I try to input these explanations into my notes. (June, Day 2)
9. My only suggestion would be to do a little less live coding but instead have the code and explain what each line is doing more thoroughly. So we can run it line by line together and take notes on our code of parameters that can be changed and how they would change things. We did this with the live coding, but with people running into errors sometimes the learning went away to fix the code instead. To do noted by Dan/Sue/Brian: let’s copy large blocks of code from lesson plan. Don’t need theme stuff (e.g., large fonts) in live coding, but do leave this in lesson plan.
10. I would suggest that the instructors copy the very long codes from the course material instead of typing them out. This would prevent participants from getting ahead of the instructor and waiting for the instructor to type out all the codes (Aug).

QC (All – for your respective sections, but particularly Antonis and Brian, including for Elaheh’s deconvolution section)
1. What things in your experience usually go wrong? What should we be checking that isn’t here? Do you have to worry about cell cycle (June, Day 1)?

Overview / motivation (Brian)
1. Can the pipeline apply to other kinds of tissues? Like kidney, lung, etc (June, Day 1).
2. More focus on or mention of biological context might help understanding the data processing steps better (June, Day 1).
3. I’d like to have a more “higher level” explanation of the data types, where is the information coming from? (Aug)

Experimental design (Sue)
1. If you have multiple biological replicates, are technical replicates still suggested? (Aug)

Normalization (Brian)
1. I think I’d benefit a bit more from revisiting the topic of normalization (June, Day 1).
2. A deeper higher-level understanding would be great for each step, i.e., differences between each normalization method. (Aug)
3. It was hard to follow the afternoon (of the first day) when the pace picked up for the heavier topics. SCTransform and normalization etc. are quite confusing. (Aug)
4. Speed was a bit (fast, presumably) for the normalization, but it came together at the end. (Aug)
5. How do you determine which method is best apart from just visualizing differential variance (June, Day 1)?
6. What should an ideal normalization look like in terms of variance, spread, etc. (June, Day 1)?
7. Why do we have a scale factor of 10^6 (June, Day 1)?
8. What was the point of finding the top 15 genes if they weren’t showing any of those nice cortical layer specific markers? If these genes weren’t location specific what were they for (June, Day 1)?

Deconvolution (Brian)
1. I was a bit confused as to why we were doing each step of deconvolution. (Aug)
2. Can we learn to define cell types without the reference? (Aug)
3. Can you only do deconvolution if you have single cell from that tissue? (Aug)
4. Was the scRNA-seq derived from the same tissue? How does this work when there is a disease stage? (Aug)
5. How to do data driven cell calling from canonical markers instead of predefined lists? (Aug)
6. Can RCTD be used in other applications? (Don’t know what this is referring to.)

Other/downstream analyses (Brian and Antonis)
1. How can seurat and ST be used to associate one “spot” with another? What kinds of maps can we make with this kind of data (June, Day 1)?
2. Can we learn how to merge / integrate many samples? (Aug)