Cells per cluster are identical, so I suspect changes to SingleR are responsible for this different output.
Summary
Small changes in the SingleR output are not affecting the summary in the text, so it's probably fine just to use the new results.
You might want to add a sanity check that the code results continue to match the text.
giving identical results on my machine to the published version, so I again suspect changes to SingleR are responsible for this different output.
Summary
Small changes in the SingleR output are giving slightly different Rand indices, but this doesn't really affect the summary in the main text, so it's probably fine just to use the new results.
You will need to modify the existing sanity check to allow for the new value of rand.
Although a seed for random number generation is set in this code chunk and prior chunks that it depends on, the vagaries of (potentially stochastic) and numerical precision of PCA/cPCA/t-SNE and/or small changes to code underlying these functions may have lead to the small differences in the t-SNE plots (beyond the simple mirroring
Summary
Small changes in code underlying the PCA/cPCA/t-SNE have resulted in changes to the t-SNE, but this doesn't really affect the summary in the main text, so it's probably fine just to use the new results.
@lgeistlinger I think this has been broken for a while: https://www.bioconductor.org/checkResults/3.18/books-LATEST/OSCA.advanced/nebbiolo2-buildsrc.html. I can reproduce the error on my Ubuntu laptop running BioC 3.18 (session info below). Stepping through
inst/book/cell-cycle.Rmd
, I find a few differences in the output to what is currently published to https://bioconductor.org/books/3.18/OSCA.advanced/cell-cycle-assignment.html. Not all differences results in errors.Hope you can decide on a fix and that my suggestions are helpful. I'm away at a conference from today and until the week after next.
Difference 1
This does not result in an error in the build.
Part A
Code chunk
https://github.com/OSCA-source/OSCA.advanced/blob/9c6309bd315d4e35d0b7e0d6390ba46508107df5/inst/book/cell-cycle.Rmd#L137-L148
Current published output
Output on my machine
Hypothesis
Cells per cluster are identical, so I suspect changes to SingleR are responsible for this different output.
Summary
Small changes in the SingleR output are not affecting the summary in the text, so it's probably fine just to use the new results. You might want to add a sanity check that the code results continue to match the text.
Part B
Code chunk
https://github.com/OSCA-source/OSCA.advanced/blob/9c6309bd315d4e35d0b7e0d6390ba46508107df5/inst/book/cell-cycle.Rmd#L183-L186
Current published output
Output on my machine
Hypothesis
This is a consequence of Part A.
Difference 2
This is the error that causes the build to fail.
Code chunk
https://github.com/OSCA-source/OSCA.advanced/blob/9c6309bd315d4e35d0b7e0d6390ba46508107df5/inst/book/cell-cycle.Rmd#L224-L233
Current published output
The results of this code chunk are hidden, but the key point is that it used to be true that
rand > 0.6
.Output on my machine
Hypothesis
cyclone-assigned cell cycle phases appear identical, based on:
https://github.com/OSCA-source/OSCA.advanced/blob/9c6309bd315d4e35d0b7e0d6390ba46508107df5/inst/book/cell-cycle.Rmd#L221
giving identical results on my machine to the published version, so I again suspect changes to SingleR are responsible for this different output.
Summary
Small changes in the SingleR output are giving slightly different Rand indices, but this doesn't really affect the summary in the main text, so it's probably fine just to use the new results. You will need to modify the existing sanity check to allow for the new value of
rand
.Difference 3
This does not result in an error in the build.
Code chunk
https://github.com/OSCA-source/OSCA.advanced/blob/9c6309bd315d4e35d0b7e0d6390ba46508107df5/inst/book/cell-cycle.Rmd#L394-L421
Current published output
Output on my machine
Hypothesis
Although a seed for random number generation is set in this code chunk and prior chunks that it depends on, the vagaries of (potentially stochastic) and numerical precision of PCA/cPCA/t-SNE and/or small changes to code underlying these functions may have lead to the small differences in the t-SNE plots (beyond the simple mirroring
Summary
Small changes in code underlying the PCA/cPCA/t-SNE have resulted in changes to the t-SNE, but this doesn't really affect the summary in the main text, so it's probably fine just to use the new results.
Session info