Story

Erika is wondering if there are different expression characteristics based on cell size. In particular, spermatocytes are much larger than other cell types; do they have higher amount of coverage, or another signature that we can use to separate cell types by size.

Let's focus on this vignette:

http://satijalab.org/seurat/pbmc3k_tutorial.html

They showed in worm that histone dynamics are changing as germ cells are moving through development.

Questions

• Can we use histone dynamics to identify different germ cell populations at different developmental stages?

Story

One thing I have worried about is the X;AA results when looking at different gene lists. We cannot simply use all expressed genes, b/c expression is so sparse in these data. To combat this I want to on "housekeeping genes", but this is challenging to define. I have looked at several different sets and see similar X;AA results no matter which set I select. Need some kind of figure depicting the X expression level using these different sets by cluster.

Questions and Tasks

• Design a figure element that shows X expression level relative to gene set.
• Select a statistic of overlap and compare different gene sets.
• Add text describing results.

Definition of done

• PDF of figure element on DGS.
• Write up of statistical comparison added to the text.

Story

Brian wants a heatmap showing differential expression arranged by cluster. This will be a major figure element showing major patterns of differential expression. Along with pulling out specific examples based on genes known in the literature.

Questions and Tasks

• Heatmap with genes (rows) and cells (columns) hierarchically clustered by rows and ordered by KNN cluster ID for columns.
• Heatmap with genes (rows) and KNN cluster (columns) hierarchically clustered rows.
• Repeat heatmaps above with with top 50 genes.
• Repeat heatmaps above with literature genes

Definition of done

• PDF versions of each heatmap placed on DGS.

From Brian:

I would like use to move away from the projections onto the tSNEs for looking at cluster gene expression. We will still have one color coded tSNE, but I think I would rather highlight genes in a heatmap or rows of histograms where we can look at each of the genes where we will use an image and each of the genes that we discuss in the text in the next part of the figure. This gets around the problem with cells from one group looking like they belong to another. I think we could do expression in the group/average in all or a z-score. I would like to look for general tTAF signatures (translational control motif, altered core promoter sequence, any no-random distribution in clusters of expressed genes, or inactivation spreading). We still need to reparse genes for looking at X expression (including t=other ways of getting housekeeping functions, and looking at highly testis-biased expression as outlined at the end of the current text. We will also want to extend this comparative analysis to look at gene movement, especially using the species with neo-Xs.

Story

To simplify the paper we want to look at the DCC and X store using median or sum expression across all somatic cells. Since our focus is not on the soma, it may be nice to focus on them all together as apposed to individual somatic clusters.

Questions and Tasks

• Combine all somatic cells and rebuild the boxplots

Definition of done

• PDF version of boxplot on DGS

Check the read lengths and make sure they look ok.

I1 == 8bp
R1 == 26bp
R2 == 98bp

To Do
[ ] Figure out what format cell ranger data needs to be in.
[ ] Reshape cell ranger data to fit into seurat
[ ] Go through vignette and write script

Story

Questions and Tasks

Definition of done

Sanity check the split reads we got from core and make sure counts add up.

To Do
[X] Read bioarxiv paper
[X] Set up meeting time with Sharvani to discuss
[X] Meet with Sharvani

Looking at cell ranger count results we see that the Testis sample has really nice clusters and the Ovaries have a few clusters. These are presumably different cell types. I would like to

1. Determine what genes make contribute to these separations
2. Determine what cell type(s) each cluster represents.

@sharvanim,

Can you compile a list of links for the different single cell seq types?

As we are going though the literature we will be running into things like 10x, SMART-Seq2, MARS-Seq, ...

It would be nice to have a list of links compiled for quick reference, and since you have done a lot more reading on the topic I figured you have a better idea of what to list and where to link to.

Need to get any emails from @sharvanim related to what was done by the core. That way I can decide if there is other information needed.

There is a hypothesis that as a germ cell differentiates the X-dosage compensation/activation would change. This could be determined by looking at X to autosome ratios.

Do we need to split the reads into cell specific files. I am sure cell ranger can take care of this for us.

For the three potential targets of Dsx, we expect basically the same thing. They should be dsx targets and the tj>RNAi experiments suggest that they should have sex-specific or sex-biased expression in tj expressing cells and that they are positively regulated by dsxF. Br requires more analysis as there are so many alternative promoters to work through. Skd is the one we have the most on, where we expect that it should be lower in dsx and tj cells in testis than in the others, while we expect high expression in the females. What I would like to see are some histograms. For example, divide the testis cells into two groups, those that express dsx and those that don’t and determine what the distribution of skd reads are in those two groups of cells. Do the same for ovaries, even though they do not cluster very well. Repeat but parse by tj+ or tj- cells. Repeat, but parse for chinmo+ or chinmo- cells. See what you come up with for differences in skd msn and br expression in the sexes.

To Do
[ ] Run samtools idxstats to get the number of reads per chromosome
[ ] Take gene counts and look at the distribution of genes by chromsome

Refine the preliminary analysis after discussing results with @sharvanim and Brian.

Story

Brian wants a tSNE with everything editable so that he can change colors to his liking.

Questions and Tasks

• Generate a PDF formated tSNE.

Definition of done

• PDF placed on DGS.

Need a list of all "biomarkers" that we can use to separate cell types. This will be most useful to trying to identify doublets.

If we look at each cell, how many genes are expressed. We know on average it was ~2.5k, are there some cells with <500 genes expressed?

To Do
[ ] Look through vignette and see if they show how to get "gene importance" for tSNE clusters.
[ ] Explore output/robject blocks to find where cluster information is stored for tSNE
[ ] Reshape output and identify key genes that contribute to clustering
[ ] Give @sharvanim list of key genes and have her figure out cluster cell type(s)

Story

Sharvani has sequenced two samples of larval gonads (testis, ovary). The goal of this preliminary study is to get a feel for scRNA-Seq data, and determine how many cells are needed to answer the types of questions we want to ask.

Questions and Tasks

• What set of scRNA-Seq tools should we use for these analysis and why?
• What is the overall quality of scRNA-Seq data that we get from the 10x genomics system?
• How much data do we need, for scRNA-Seq to be useful?

Definition of Done

• A scRNA-Seq workflow.
• A plan on how to proceed with future scRNA-Seq experiments.
• Preliminary analysis looking at germline soma communication.

I want to know what people are currently doing in single-cell analysis.

Need to look into the math of a tsne and see what they are really representing. Does cluster structure really have biological meaning, or is it a mathmatical massage.

To Do
[ ] Run seurat on Ovary sample
[ ] Run seurat on Testis sample

Story

We are concerned about contamination by dead/dying cells. Are there specific transcriptional signatures that we can use to separate cells that look dead. Erika pointed out that germ cells don't use apoptosis, so there may need to be other signatures.