[NCP Stem 1] Profile 22q cohort stem cells (D0)

[mtegtmey]

Goal

• Perform Cell Painting on iPSCs, akin to cmQTL project

Experimental Design

Expected date for imaging: Done
Dyes: Cell Painting dyes
Cell type: Day 0 stem cells
Plates: 1x 384-well
Plate layout: this will be identical to the layout used for the cmQTL project, consisting of 48 different lines segmented into 4-well blocks dispersed across the 384-well plate.
Plating parameters: 10k cells per well fixed 6hrs post-plating (same as cmQTL)

Proposed analysis:

1. Ensure we can stratify samples
2. Identify particular features and organelles structures perturbed by the 22q11 deletion
3. Using existing RNA-expression data to integrate imaging and molecular data

Metadata

• 22q11.2_cellpaint_iPSCs_metadata.xlsx

[shntnu]

I need to rename before uploading; will do

ls -1 /imaging/analysis/2019_05_28_Neuronal_Cell_Painting/NCP_STEM_1/images
Assay Plate 0_1 ??? Measurement 219___2020-09-03T16_51_45-Measurement 1
Assay Plate 0_1__2020-09-02T17_40_54-Measurement 219

[mtegtmey]

@shntnu @gwaygenomics Hi guys, i wanted to touch base about any updates in terms of analysis of the iPSC data from this project?

I am gearing up to have the neuronal progenitor (NPC) plates screened in about two weeks, before the holiday.

I am also trying to plan a way in which, when doing the NPC plate, I can continue some of these cells into D28 neurons to be screened later - how confident are we about having some of the neuronal cell type parameters nailed down within the next month?

[shntnu]

@mtegtmey this is on my plate, still pending. I'll peek in tomorrow and give you an estimate of what's remaining

[shntnu]

NCP Stem 1 is finally underway! #9

For our notes, I used this CellProfiler pipeline
https://github.com/broadinstitute/imaging-platform-pipelines/tree/master/cellpainting_ipsc_20x_phenix_with_bf_bin1_cp3

(this is documented here https://github.com/broadinstitute/neuronal-cell-painting/blob/3ada6f60643f8d3659d8c27f55e28a05987d739d/1.main-run-workflows/generate_profiles.sh#L73)

[shntnu]

Checking platemap

@mtegtmey Do these look right?

[mtegtmey]

@shntnu yep! These looks good to me. I hope you're also able to enjoy your weekend :)

[shntnu]

Great! I'll proceed with this :)

[shntnu]

@mtegtmey @raldanehme

I repurposed a cmQTL notebook to inspect the profiles
https://github.com/broadinstitute/cmQTL/blob/master/1.profile-cell-lines/4.inspect-profiles.md

to produce
https://github.com/broadinstitute/neuronal-cell-painting/blob/master/1.main-run-workflows/0.inspect-profiles.md

At a high level, the replicate correlations look great, so that's a relief!

I don't know how much to read into about this plot

The relationship between cell count and replicate correlation is not much so I don't think it matters too much

Have a look, and we can next start pondering whether we should do some early analysis on this or wait for NCP_Progenitors_1 before delving deeper.

---

@gwaygenomics I reverted to the old profiling workflow to keep things moving (we were thinking we'd wait for Niranj to push to https://github.com/cytomining/profiling-recipe but that's going to take a few more weeks.

[mtegtmey]

@shntnu I think the general trend of higher # IDs having higher cell counts is observed in the cmQTL project more broadly. It's likely driven by the time it takes to aliquot all of the samples compared with the resuspension of cells prior to using the robot. I don't think it means much otherwise - though it is good to see that distribution isn't as varied as we have seen previously and the overall cell counts per sample are pretty good.

Given the holiday, it's looking like data won't be generated on the NPC plate until the week of Nov 30 - Dec 4, im still waiting to hear back from Masha about her schedule for next week, and the staining team. Will update you!

If possible, moving forward with an early analysis would be ideal. I have a committee meeting coming in late Jan and this data could inform some of the experiments I'd like to do before then. However, if it's simpler for you do do the analysis in bulk, waiting an 1-2 weeks for the NPC data certainly won't be a barrier for me! :)

[shntnu]

I think the general trend of higher # IDs having higher cell counts is observed in the cmQTL project more broadly. It's likely driven by the time it takes to aliquot all of the samples compared with the resuspension of cells prior to using the robot. I don't think it means much otherwise - though it is good to see that distribution isn't as varied as we have seen previously and the overall cell counts per sample are pretty good.

Agreed

If possible, moving forward with an early analysis would be ideal. I have a committee meeting coming in late Jan and this data could inform some of the experiments I'd like to do before then. However, if it's simpler for you do do the analysis in bulk, waiting an 1-2 weeks for the NPC data certainly won't be a barrier for me! :)

Sounds good. I'll prioritize getting the other two plates to the same stage as this one and then figure out the next steps together.

[mtegtmey]

@shntnu any particular place i can upload the images for NCP NPC 1?

[shntnu]

• Identify particular features and organelles structures perturbed by the 22q11 deletion

For this goal, I will use @gwaygenomics 's code here as the starting point
https://github.com/broadinstitute/profiling-resistance-mechanisms/blob/master/3.bulk-signatures/1.derive-bulk-signatures.ipynb

[ruifanp]

Spearman correlations within each human/isogenic sample ID.

@shntnu does this seem consistent with past results?

[shntnu]

@shntnu does this seem consistent with past results?

The correlations are generally high, and more variable for the corhort cell lines, and less variable for the isogenic; this is definitely consistent.

[ruifanp]

I compared the default cellprofiler feature selection with the feature selection method I had previously used for the cardiomyocyte project. Additionally, I tested the effect of reversing the labels in the isogenic samples, in case they were flipped in the first place.

code: https://github.com/ruifanp/neuronal-cell-painting/blob/master/analysis/STEM1_01_eda.ipynb

The number of statistically significant features were features which were found to differ between control and deletion in both human and isogenic samples. Using my feature selection appears to improve the data signal by improving replicate correlation, number of significant features (suggesting that my features were more informative), and cluster homogeneity. My feature selection results in a stronger signal and suggests that the isogenic labels might have been flipped. Otherwise, it wouldn't make sense if the features in agreement between human and isogenic have a lower occurrence in the statistically significant features vs overall.

[shntnu]

Thanks @ruifanp. Let's ponder this swap during tomorrow's profiling check-in.

[shntnu]

@raldanehme @mtegtmey Now that we've sorted out the staining plan for neurons (#6 and #8), I assume we should now get focus on the analysis of 22q stem cells? @ruifanp is already doing this but wanted to make sure we'd rather do this than work on the progenitor 22q dataset.

[mtegtmey]

@shntnu @ruifanp personally, I like to move forward with the NPC data. This would be key for my PhD related progress. But if ‘finishing’ up the stem cell data is relatively quick, definitely move ahead with that. Any data suggesting feature differences in the presence of the deletion will be very helpful for me, regardless of cell type!

…

On Jun 24, 2021, at 7:23 AM, Shantanu Singh ***@***.***> wrote:  @raldanehme @mtegtmey Now that we've sorted out the staining plan for neurons (#6 and #8), I assume we should now get focus on the analysis of 22q stem cells? @ruifanp is already doing this but wanted to make sure we'd rather do this than work on the progenitor 22q dataset. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[shntnu]

@mtegtmey We can certainly prioritize that. Ruifan's analysis will be equally applicable and it may even help him to have two datasets to work with simultaneously. It might be confusing to keep two threads active (this one and #10) but will keep them both open for now. I'll ping you on the other one with some questions.

[shntnu]

@yhan8 For your work on this project (using DL-based features to start with), we will likely use this dataset (i.e. Day 0).

@mtegtmey – @yhan8 (whom you met a few weeks ago IIRC) will be joining the project and to get started, one of her goals is to extract deep learning-based features on these data and reproduce the same analysis that @ruifanp performed on these datasets. We will discuss longer-term goals the next time we connect. Meanwhile @ruifanp is continuing his analysis on progenitors here #10 (the results from today look very promising)

[mtegtmey]

Exciting stuff! Welcome aboard @yhan8

…

On Oct 13, 2021, at 1:41 PM, Shantanu Singh ***@***.***> wrote: @yhan8 <https://github.com/yhan8> For your work on this project (using DL-based features to start with), we will likely use this dataset (i.e. Day 0). @mtegtmey <https://github.com/mtegtmey> – @yhan8 <https://github.com/yhan8> (whom you met a few weeks ago IIRC) will be joining the project and to get started, one of her goals is to extract deep learning-based features on these data and reproduce the same analysis that @ruifanp <https://github.com/ruifanp> performed on these datasets. We will discuss longer-term goals the next time we connect. Meanwhile @ruifanp <https://github.com/ruifanp> is continuing his analysis on progenitors here #10 <#10> (the results from today look very promising) — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#7 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AMSE5EROBBSIWTCZJ6AE5RDUGXAF3ANCNFSM4RYYD54Q>.