Eccentric medium spiny neuron (eMSN)

[jnmaciuch]

I think there might be a misunderstanding because I assume these were genes were taken from Paolo's supplementary table 5? I don't think these are the genes driving the eMSN category signal. They are just the ones that were in that category and also in the list of 115 genes from the Zhang et al. preprint.

I don't think that's true because it comes from the columns labeled "GeneRatio", "BgRatio", "FoldEnrichment", "genes" and the number of genes in that last column is the same as the numerator as the GeneRatio column. That's standard output for an overenrichment analysis. Plus the methods detail creating those cell-type specific gene lists and as far as I can tell from the data that's the only place they would have been used. Also not all the leading edge genes in the list were in Zhang et al.

The eMSN category consists of hundreds of genes and the signal isn't driven by handful of them but spread out across many weak signals for many genes.

Yes it's a leading edge (though it looks like the cell type gene lists generated had a few dozen genes, looking at the denominator of GeneRatio). They would've come from the part of the methods setting a logFC cutoff for expression between that cell type and the average for all others

 

[forestglip]

I don't think that's true because it comes from the columns labeled "GeneRatio", "BgRatio", "FoldEnrichment", "genes" and the number of genes in that last column is the same as the numerator as the GeneRatio column. That's standard output for an overenrichment analysis

GeneRatio is defined as the following in the Readme of the supplementary file, so I think it's Paolo's own metric:

number of HEAL2 genes in the gene set/number of HEAL2 genes in background

Also not all the leading edge genes in the list were in Zhang et al.

I think that BAIAP2, the one you said isn't in the list, also is in there at position 107.

 

[jnmaciuch]

I think that BAIAP2, the one you said isn't in the list, also is in there.

Ah okay I stand corrected. So it's overrepresentation analysis of HEAL2, that makes more sense why the expression plots diverged so much

[Edit: That part of the methods would be for generating the background gene list per cell type and brain region, then]

 

[jnmaciuch]

So I guess my quick and sloppy sanity check just points to the fact that [Edit: many of the top Zhang et al. hits point more to MSNs, and overlap with eMSN signal in DecodeME seems to be more driven by similarity between MSN/eMSN]