search

jhsiao999 / peco

Predict cell cycle phase in a continuum from single-cell RNA-seq data
https://jhsiao999.github.io/peco/
GNU General Public License v3.0
12 stars 4 forks source link

prediction comments #42

Open kasperdanielhansen opened 4 years ago

kasperdanielhansen commented 4 years ago

1) The vignette says the predictor uses 5 genes, as does the paper, but this is hard to see from the code. If true, I suggest you to be more clear on this.

2) You apply the predictor to your own training data set (I think). While useful, the call to cycle_npreg_outsample makes it a bit unclear what the requirement is for prediction. For example, should it have the same row names, have the same gene ordering etc?

pcarbo commented 3 years ago

@jhsiao999 If I understand the vignette correctly, the example uses 101 genes to predict cell cycle, not 5. I do agree with @kasperdanielhansen it is a bit confusing because you talk about only needing 5 genes at the beginning of the vignette.

I also agree that it would be helpful if help(cycle_npreg_outsample) would give more information about the Y_test input argument. Currently it just says this:

Y_test: A ‘SingleCellExperiment’ object.

In particular, it is unclear how the Y_test input is supposed to align with the Y input to cycle_npreg_insample.

pcarbo commented 3 years ago

@kasperdanielhansen Note there is also an example here that might help because it illustrates the use of both cycle_npreg_insample and cycle_npreg_outsample.