input data normalisation

[bjstewart1]

Can you clarify what preprocessing is expected for the RNA assay (gene expression) data.
Is the expected input integer (raw) counts or normalised&log transformed counts or similar?

this isn't totally clear in the vignette.

[zhenzuo2]

I also have a similar question. I am not sure if the returned seurat object contaions the data used for genearlized liear regression, usch as vecotor from RNA assay, or ATAC assay.

[zhenzuo2]

I also have a similar question. I am not sure if the returned seurat object contaions the data used for genearlized liear regression, usch as vecotor from RNA assay, or ATAC assay.

Here is an example. I can access normalized expression level of PAX6 but I am not sure how to access value of "chr1-911275-911316".

[zhenzuo2]

I also have a similar question. I am not sure if the returned seurat object contaions the data used for genearlized liear regression, usch as vecotor from RNA assay, or ATAC assay.

Here is an example. I can access normalized expression level of PAX6 but I am not sure how to access value of "chr1-911275-911316".

@bjstewart1 Using this as an exmaple, you question is what is the values for PAX6. Are they norlized values, scaled values, or raw counts.

[bjstewart1]

I also have a similar question. I am not sure if the returned seurat object contaions the data used for genearlized liear regression, usch as vecotor from RNA assay, or ATAC assay.

Here is an example. I can access normalized expression level of PAX6 but I am not sure how to access value of "chr1-911275-911316".

@bjstewart1 Using this as an exmaple, you question is what is the values for PAX6. Are they norlized values, scaled values, or raw counts.

no my question is what are the input data for the tool. Are the RNA integer counts meant to be processed to normalised/log transformed values?

[joschif]

Hi @bjstewart1, currently Pando would expect log-normalized data (for RNA) and tf-idf-normalized data (for ATAC) as input and would also generally use that by default if it's in the data slot of your assay. I've thought about implementing an option to run it on raw counts though - essentially that would require other noise models for the GLMs and accounting for library size covariates in the models.

[bjstewart1]

Thanks @joschif really helpful .. - can I suggest that you make it a bit clearer what these input requirements are in the readme/vignettes?

[elhaam]

Hi @joschif ,

Thanks for your response above! I have a follow-up question. When pre-processing data, do you suggest standard QC and filtering (for example min.cells = 3, min.features = 200) in RNA-Seq? I believe this tutorial has not performed QC steps to filter out genes since I see ~31k genes for RNA data. Could you please clarify if we need to keep all the genes, then normalize and get log1p?

Thanks,
Elham