[NCP Progenitors 1] Adding branching metrics

[gwaybio]

We added branching measurements in NCP Pilot 3B (see #16) and, by eye, they appeared useful.

We will also add these measurements to the NCP Progenitors 1 dataset

[pearlryder]

Hi everyone,

I've taken a look at several images from the Progenitors dataset to assess the feasibility of a branching analysis.

First, I remembered that there was heterogeneity across the different wells regarding confluency and that the more confluent cells did not extend branches. I wanted to check to see if this was related to the biological condition (controls vs patient samples, isogenic control vs deletion). If I understand the experimental design correctly, all wells were plated w/ 15k cells (#10), so that is not a factor in this case.

Does genetic condition affect confluency?
I checked 10 wells from across the plate for the control line vs deletion line. I was using the metadata map for NCP Stem 1. I found that some wells were very confluent and some much more sparse in both the control and deletion line. I checked two wells each from the isogenic control and deletion lines and found that they were about the same confluency.

Here are a few example images:
Controls:

Deletion:

Conclusion: My interpretation is that confluency at least is not a barrier to proceeding with a branching analysis. If all of the controls were very confluent and all of the deletions were very sparse, I think that would make any interpretation of results pretty difficult (since the confluent cells don't have any processes extended). But we seem to have a range of confluencies present in both biological conditions. A caveat to this analysis will be the presence of very confluent cells that essentially don't demonstrate any branching.

Should I adopt our segmentation to better identify processes?
This question is mostly for @bethac07 and @shntnu. If you see the images below, my initial segmentation method didn't capture the dim extensions that extend from some cells.

Should I try to adjust the segmentation parameters to better capture these processes before we add the branching measurements?

• This should be possible, it will just require about an hour of time to adapt the pipeline
• @shntnu, have you already analyzed the previous dataset extensively? There seems to be potential here to have multiple sets of metrics from the same images, which could get confusing. If I do re-segment the cells, should I replace the previous metrics or create a new database?

Thanks in advance to everyone for the feedback!

[bethac07]

I'll punt on "should those be captured" to @mtegtmey and @raldanehme ; since they are SO much dimmer, my feeling would be "no", but I'm not as familiar with the biology here.

[mtegtmey]

@pearlryder thanks for looking into this! We are concerned that adapting the pipeline would alter Shantanu's downstream analysis, as those metrics are more important to us to define their morphological profiles. That said, the 'branching' of these cells is of significant interest to us and we expect there to be a difference between conditions here.

Would it be possible to adjust the segmentation parameters and only capture the branching metric, ignoring all the others? Rather than adjusting segmentation and repeating the downstream analysis for the other features. Please let me know if I'm thinking about this incorrectly!

[pearlryder]

Thanks @mtegtmey! That sounds like a reasonable approach to us. I'll go ahead and adapt our segmentation to better capture those dim processes, but I will use these new segmentations for branching metrics only. Please let us know if you have any questions!

[pearlryder]

Hi all,

The branching analysis for the NCP progenitors experiment has been completed in CellProfiler. @shntnu, the backends are available at s3://imaging-platform/projects/2019_05_28_Neuronal_Cell_Painting/workspace/backend/NCP_PROGENITORS_1_BRANCHING/. Let me know if there are any questions!

[shntnu]

Note for ourselves that the branching analysis essentially generates four extra features (*_CellBodiesPlusNeuritesSkel); the full list is below, from here.

variables
Cells_Children_CellBodiesPlusNeurites_Count
Cells_Children_Cytoplasm_Count
Cells_Children_Neurites_Count
Cells_Location_Center_X
Cells_Location_Center_Y
Cells_Location_Center_Z
Cells_Number_Object_Number
Cells_Parent_Nuclei
Cytoplasm_Location_Center_X
Cytoplasm_Location_Center_Y
Cytoplasm_Number_Object_Number
Cytoplasm_Parent_Cells
Cytoplasm_Parent_Nuclei
Nuclei_Children_Cells_Count
Nuclei_Children_Cytoplasm_Count
Nuclei_Location_Center_X
Nuclei_Location_Center_Y
Nuclei_Location_Center_Z
Nuclei_Number_Object_Number
Nuclei_ObjectSkeleton_NumberBranchEnds_CellBodiesPlusNeuritesSkel
Nuclei_ObjectSkeleton_NumberNonTrunkBranches_CellBodiesPlusNeuritesSkel
Nuclei_ObjectSkeleton_NumberTrunks_CellBodiesPlusNeuritesSkel
Nuclei_ObjectSkeleton_TotalObjectSkeletonLength_CellBodiesPlusNeuritesSkel

[shntnu]

I'll close this issue in favor of continuing the analysis in #10