Tasks 2025

[adRn-s]

all days / general

• prepare certificate of attendance
• can we have S3 or S4 taught in earlier courses? method dispatch comes up with ?RunUMAP and n.dims being only defined in the latter listing (dispatch for SeuratObj). We insist into looking at help pages but this requires some clarification (so far, improvised/ spontaneus)
• add slides to repo (editable to all of us. Either PPT or Quarto, not PDF.)
• how can we encourage more interaction? polls put the students to work, but they don't discuss or talk too much.
  • shall we convert some polls to exercises? and have students share their answers after we walk among them seeing what they're doing (interact more la 1:1, so that we know where they are, we were lacking that first hand feedback on what's understood, what's overlooked during explanations, and so on..)
• clean sessionInfo, it's the same environment everywhere. Let's have it only in the PkgInstall doc.
• centralize poll answers in this repo, we had mismatches in the past (wiki, pollev website, and Rmd)
  • remove any answers or "proposed solution" from Rmd(s)
  • try adding solutions in file(s) under rmd_archive since this folder is hardly seen by students.
  • have a CI job to run these solutions to exercises too (but not add to website)
  • add instructions for instructors, how to upload csv, and set the "disallowing" of changing answers by respondents.
• check Tasks 2024 #115 all are done and removed, but there's some stuff there still that remains to be reviewed and possibly added to this repo itself (e.g. README html comment)
• move these tasks, or whatever would be relevant to future iterations, to a file within the repo (README?)
• regarding our analysis extensions, shall we add...?
  • https://rsconnect.ie-freiburg.mpg.de/BioinfoShinyApps/CellViz/
  • iSEE (for QC)

PkgInstall.Rmd

• we forgot seuratdata, hdf5r and presto. Also, MatrixGenerics had to be backported and even then MAST was still broken for SCT
• the package installation script might scare many students. we need to decide between:
  • do it with students on day 1? it should take us ~15 minutes.
  • use .libPaths
• use tabsets for all the "skip" and "not to skip" bifurcation (conflict: quarto don't have these... if we wanted to migrate in the future)

day1

• we need to insist on the importance of the end-of-course survey from day 1, so that everyone does it no matter if they dropped out. this feedback is utilized, and we depend upon their availability to guide us and improve for the future generations of students!
• we should also mention the course programme on day 1, and make it clear that the last "DIY" day is the most important, according to past students (surveys), it's the day that everything 'clicks'.
• add some exercises with accessor functions and get more intimate with SeuratObj. Check essential commands html. It should be covered on a high proportion.
• add a "head matrix.mtx" of sorts before loading the data, so that they see the plaintext file content.
• shall we move KNN, and UMAP, to day 2? It would allow us to expand a bit more on transformations, too.
• first exercise should be a histogram of another gene, so that zero inflation is self-evident. Or try custom bin size/ density plot?...
• remove links to PollEv, use the shortened 'thomasmanke101' instead (or QR).
• fix CSV and PollEv answers to "Poll: How many cells have equal or less than 2000 counts in total (summed over all genes)?", the wiki holds the correct version.
• add some explanation o the method=vst (and mvp?) of FindVariableFeatures since its just a linear model.
  • can we do this manually with tidymodels and pbmc.data matrix?
  • explore the number (2000) as parameter and effect on downstream (eg. Elbowplot)
• recap polls from day2 could do a lot better. Simpler answers would help. edit: DONE, moved them over to this day, and instead of polling with 'tricky' options, let's just give away the answers (hinting towards simple ML gotchas)

day2

• intro polls to recap day1 could do a lot better if we move them as exercises in that day
• on the new 20.Rmd: add some exercises, for example use RunUMAP without FindNeighbors
• de: last option on 2nd-to-last question should be removed, and we should accept the first one as correct: umap is independent of clustering. (don't they share the shared-nearest neighbor graph?)
• de: last question had weird options. options 1 and 2 are included in the last. And, what's a gene signature? we need to explain a bit there.
• batch effects, poll 2nd question was missing on PollEV. I just edited Rmd, and added it to CSV. Also, add answer to wiki?
• disentangle sctransform from integration and DE (as it currently stands). Instead, just add a warning at the end. Oh, btw, there are these functions (prep-anchors and prep-markers) that will be required.
• last exercise, identifying the clusters should also hint into using dotplot together with featureplot

day3

• we had free time, even after moving enrichment analysis here. we need to add more exercises. or shift things from the previous day (better)
• findcoservedmarkers (suggested at the end of DE w/ sctransform) is not working. It requires to run seuratobj[["RNA"]] %<>% JoinLayers() according to the error message, but this didn't work with "SCT", so...

day4

• heatmap w/ counts don't really tell a lot to compare clusters ? -> clustree (?)
• dotplot doesnt work with the annotations (Clusters) because of special characters, but it works with seurat_clusters if we switch Idents.
• the "^MT-" pattern doesnt work with this data (mice), use lowercase.
• MAST was broken, maybe because MatrixGenerix had to be backported from Bioc 3.18

[adRn-s]

COURSE PLAN

Courseday	Weekday	Date	Total time	Venue
Day1	Monday	?	13:00-16:30	Lecture Hall
Day2	Tuesday	?	14:00-17:00	Lecture Hall
Day3	Wednesday	?	13:00-16:30	Lecture Hall
Day4	Thursday	?	13:00 -16:30	Lecture Hall

Courseday	Timeslot	Module	Assignment
Day1	13:00-13:30	setup	All
Day1	13:30-14:00	Intro + Parallelization	?
Day1	14:00-16:30	First steps	?
--	--	--	--
Day2	14:00-15:30	Clustering + Embeddings	?
Day2	15:30-17:00	Data Viz + DE + Ontologies	?
--	--	--	--
Day3	13:00-16:30	Batch Effects + SCTransform	?
--	--	--	--
Day4	13:00-16:00	Hands-On	All