Gene view table enhancement

AnnekathrinSilvia / magnetique

Source code for the magnetique application

MIT License

2 stars 3 forks source link

Gene view table enhancement #11

Closed CDieterich closed 2 years ago

CDieterich commented 2 years ago

I think we should include 2 columns for DTU in this table here. In the end, it would be: DGE_log2FoldChange, DGE_padj, DTU_effect size (if applicable), DTU_padj

AnnekathrinSilvia commented 2 years ago

@tbrittoborges could you tell me where I find the effect size in the SummarizedExperiment? Then I can add the two columns.

tbrittoborges commented 2 years ago

@AnnekathrinSilvia You can use the likelihood ratio statistics column (lr) as an effect size estimate for now. But later, I would prefer to have a maximum difference in transcript usage, but it needs to be computed.

so for, DCM_vs_NFD:

rowData(se)[['DRIMSeq_DCM_vs_NFD']][c('lr', 'adj_pvalue')]

gets you the effect size and the test statistic.

tbrittoborges commented 2 years ago

Note: lr is not so good because it's not as intuitive as the difference in usage.

AnnekathrinSilvia commented 2 years ago

Okay, then I'll take the lr column until we have a calculate maximum difference in the se.

tbrittoborges commented 2 years ago

Something I had in mind was having child rows (https://rstudio.github.io/DT/002-rowdetails.html) for the DTU/Carnival information, if available. But it may be complicate. So let's discuss it later.

AnnekathrinSilvia commented 2 years ago

I just pushed a version with the extended table to the I4L branch. Unfortunately, the extraction and preparation of the DTU data makes the table extremly slow. @tbrittoborges maybe we should think of a better way to preprocess the data e.g. combining it more so we do not need that many processing steps in the actual app.

tbrittoborges commented 2 years ago

It looks good. I can pre-process the data. I will also split it by contrast as you have done for the DGE data.

federicomarini commented 2 years ago

Adding a note here on what we can additionally do:

[ ] Gene symbols if found as NA could be simply repeat the Ensembl identifier?
[ ] the gene symbols can be instead of simple text a link/button that points to the NCBI gene database
[ ] genes could trigger the generation of HTML content, see e.g. https://github.com/federicomarini/GeneTonic/blob/334cc30103ec1a31e98b86da3f1cd8e3303a90f1/R/GeneTonic-extras.R#L293

As a thought: maybe the usage of some collapsible elements could speed up the rendering - instead of handling 5 pieces of output, shiny would have to deal only with the visible ones, and each additional one is handled when the user opens up the collapsible?