shbrief / genomicsupersignature Goto Github PK

Hi, I find that using the codes given in the readme file, the installation process of GSSP will fail.

The error looks like this. Could you please help me resolve it? Thanks a lot!

Hi,

Is it possible to create new models based on GenomicSuperSignature? Let's say for example, using GTEx as and input and different prior knowledge.

Hi, I tried to run the codes related to RAV model building and met errors related to package installation. My R version is 4.1.3.

For the GF package, I cannot install it based on my R. Therefore, could you please share the version of your software? Thanks a lot.

• Spearman
• other methods from dist
• straight function

Add a 'droplist' argument to drawWordcloud() function, which removes most and least common MeSH terms in the universe to avoid 'outliers' skew the word cloud.

e.g. droplist = c("Human", "RNA sequencing", ..., "Publication", "Utah")

Something like:

• https://www.r-bloggers.com/2015/09/passing-arguments-to-an-r-script-from-command-lines/

Hi, I found that this step took me a pretty long time to run:

I waited for more than one hour with one 30 GB GPU from an HPC. Is it normal? Thanks a lot.

Consider using irlba for very large matrices, or allowing use of pre-computed PCA from other packages as input to 1validate()1.

Currently, looking up which studies contributed to a RAV requires something like the following:

> ravc2 <- GenomicSuperSignature::getModel(prior = "C2")
> colData(ravc2)[colData(ravc2)$RAV == "RAV272", "studies"] 
$Cl4764_272
[1] "ERP020977" "SRP039361" "SRP045352"

And then searching for the accession numbers e.g. in the European Nucleotide Browser. And it's more difficult to find the PMIDs of studies contributing to the RAV.

It would be very convenient to have a function that outputs these directly, either with a message saying to search them in the European Nucleotide Browser or giving a direct link (e.g. https://www.ebi.ac.uk/ena/browser/view/ERP020977 for the first ERP above, although ENA browser annoyingly takes you down to the "reads" section of the page). Being able to browse PMIDs of the studies would also be very useful.

It's not clear from ?validate whether users should normalize their data in advance. As I understand from the code, no normalization is done except for z-score if scale = TRUE, so users should do a log(x+1) transformation?