Error with matrix input: select_genes() expects genes in rows, psupertime() requires genes in columns

[dpcook]

Hey @wmacnair--fantastic tool! I started experimenting with it a bit yesterday and it's been working really well with my data.

I noticed some issues when using a simple expression matrix as an input. It seems that the main function wants genes to be in the columns of a matrix input (make_x_data() transposes SCE input, but not plain matrix), but then everything wrapped in with select_genes() uses a matrix with genes in the rows (eg. calc_expressed_genes() summarizes expression with rowMeans()).

Easiest fix may be simply to require a matrix input to have genes in rows and then change make_x_data() to x_data = t(x). This is probably more intuitive for single-cell analysis in R, as most tools (Seurat, SCE) keep genes in the rows anyway.

David

[wmacnair]

Hey @dpcook - exciting that people are using it! Thanks so much for trying it out, and raising the issues.

I think my original motivation for this choice was that cells * genes makes most sense in terms of the maths. But I think you're right - the inconsistency is unhelpful, and the matrices would most likely arrive in genes * cells format anyway. I'll change the expression matrix input requirements.

In case you're interested, my roadmap for future psupertime work includes some stuff which would help address this:

• implement a psupertime object as an extension to a SingleCellExperiment object
• add unit tests...
• include all the manuscript analyses in the psupplementary

Will