ebecht
commented
2 years ago Hi @GarrettCollett Your questions are perfectly valid.
For the error related to lm.fit, this is due to background correction. I'm hoping to provide users some flexibility on that aspect in a future release. This step does not make sense if you do not have isotype controls, but if you have an autofluorescence control it should do the job. Otherwise, for now you may just enter the argument isotype = rep(targets[1], length(targets)) when calling the infinity_flow() function, and ignore the output of the background correction.
As for your first question on the number of events, with these parameters your results should be 20k * number of input files (provided they all have >40k events). It sounds like the pipeline is actually using a heavily downsampled version of your dataset. Could you make sure that the files in path_to_fcs indeed have a high-enough number of events ?
Also, did you specify a custom path_to_intermediary_files argument or did you leave the default value ? If you used a custom one, could you delete the files in /path_to_intermediary_files/subsetted_fcs/ ?
Hopefully these will help you solve your issues ! Best, Etienne
Hi Etienne, Hope you're doing well. To preface: I'm very new to R so please forgive me if my questions are elementary. I'm trying to use your package to analyze a "Minfinity Flow" dataset of 13 markers on the Cytek Aurora. No isotype control was included in this experiment. I was able to generate the UMAP and augmented .fcs files. However, the augmented .fcs files are only 500 events each. Input_downsampling and prediction_downsampling are set at 100k and 20k, respectively. I also received the following error, which I'm unsure the cause of: Error in lm.fit(x, y, offset = offset, singular.ok = singular.ok, ...) : 0 (non-NA) cases Is there any advice or recommendation you can give to boost the number of predicted events? Thanks very much, Garrett Collett