Firstly thanks for adding the transient option!

However, when response_type="transient" i.e. no genes are designated switching svd throws an error:

library(ouija)
data(example_gex)
oui <- ouija(example_gex, response_type="transient")
Error in svd(x, nu = 0, nv = k) : a dimension is zero

Presumably this is due to attempting prcomp on no genes

Perhaps just do a pca of the full Y?

Also in the readme the data is called synth_gex rather than example_gex.

Hi Kieran, thanks for the cool package. I am interested in learning more about Bayesian stuff so your other work seem interesting as well!

Recently there has been talk about how UMI count data in scRNA-seq is not zero-inflated. Instead it is recommended to model UMI counts using a negative binomial (or even a Poission) distribution. (I can share some papers if you'd like)

For this reason I was wondering if there was a way to omit the explicit modeling of zero counts. Also your thoughts on using the aforementioned distributions to directly use the gene counts instead of the log-transformed CPM data.

Thanks!

Hello,
thanks for developing this tool! I am very eager to try it out.

I am having the problem that it seem like I am "loosing" the gene information in the ouija fit, but calling ouija works in priniple.

I am working on a sce, where I still see my gene names I am subsetting for:

SCEsmart[smart.top[1:10],]
class: SingleCellExperiment
dim: 10 94
metadata(1): log.exprs.offset
assays(2): counts logcounts
rownames(10): Jam2 Lgals3 ... Tcl1 Cthrc1
rowData names(12): ENSEMBL SYMBOL ... log10_total_counts AveCount
colnames(94): X00h_01_S24947 X00h_02_S24948 ... X48h_33_S25022 X48h_34_S25023
colData names(39): cellID timepoint ... pct_counts_in_top_500_features_Mito phases
reducedDimNames(0):
spikeNames(0):

options(mc.cores = parallel::detectCores())
oui <- ouija(SCEsmart[smart.top[1:10],],
single_cell_experiment_assay = "logcounts" ,
inference_type = "hmc")

print(oui)
A Ouija fit with 94 cells and 10 marker genes
Inference type: Hamiltonian Monte Carlo
MCMC info: 10000 iterations on 1 chains
(Gene behaviour) Switch/transient: 10 / 0

plot_expression(oui)

so here I do not see the genenames anymore. Any downstream analysis produces errors like:

plot_switch_times(oui)
Unknown or uninitialised column: 'Gene'.Unknown or uninitialised column: 'Gene'.
Error in $<-.data.frame(*tmp*, "Gene", value = integer(0)) : replacement has 0 rows, data has 10

gene_regs <- gene_regulation(oui)
Error: All columns in a tibble must be 1d or 2d objects: * Column gene_i is NULL * Column gene_j is NULL Call rlang::last_error() to see a backtrace

I have to confess that my understanding of the under laying data structure is limited. Do you have any suggestions what could cause this behaviour and how I could fix it?

Thank you for any help!
Best,
Merrit

Initialise all parameters from PCA etc to save MCMC runs

Hi Kieran,
Thanks for developing ouija. I'm testing it out on a complete 1386 cell x 28000 gene matrix of single cell RNASeq counts. I tested 200GB-1T memory and 1-4 CPUs. It seems to use a steady 600GB memory and cycles between 1 and 2 CPU. With the code from the readme, it has not finished (converged?) in over a day. Is the matrix too big for ouija?

I also notice there are lots of warnings before it says "SAMPLING FOR MODEL 'ouija' NOW (CHAIN 1)", and no other notices beyond that. Thanks for your advice.

library(ouija)
library(Seurat)
load("Seurat.Object.RData")
options(mc.cores = parallel::detectCores())
oui <- ouija(as.matrix(seurat@data))

Hello,
Can I fit the model and find the switch-like genes based on existing pseudotime defined by other methods?
Thanks!

As it stands the documentation for the main function is pretty poor. In particular -

• The title is "Fit a Ouija object." - could be more informative
• The description is totally non-informative
• The "return" section is close to useless
• The description should contain a section on how to control inference - this can be taken from the main vignette
• The description should at least reference the downstream analysis functions or reference to the relevant vignette section
• The arguments sent to rstan should be documented and defaults provided!

Hi,
I am new to Ouija but quite excite about using it.
I have got my Seurat object transformed into a singleCellExperiment object. Unfortunately Ouija doesn't seem to see like this. I get this :

oui <- ouija(sample1.sce)
Error in t.default(x@assays[[single_cell_experiment_assay]]) : 
  argument is not a matrix

Am I doing something wrong?

Exciting to try this out!

I didn't have any joy though, applying ouija to an SCESet object (see attached pic).

Looks like ouija could be clobbering the scater NAMESPACE on load? Or trying to access exprs() from scater when it actually comes from BiocGenerics/Biobase?

Let me know if you can't replicate - I'm using scater devel at the moment.

Please add a LICENSE (or LICENSE.md or LICENSE.txt) file to the repository, ideally using an OSI https://opensource.org/ approved open source licence compatible with the wider R Bioinformatics community to maximise uptake and reused. e.g. MIT license.

Apologies in advance as I realise this is partially a clone of #5. But I'm looking for several of the functionalities of ouija, but with a precomputed pseudotime from another pseudotime method. In particular, the plot_switch_times (with the error bounds) as well as the gene regulation and metastable state testing would be interesting. But I haven't been able to find similar functionality in switchde. Is there a good way to do this using either package?

Is it possible to plot the cells as a function of a list of genes?