Make readable circos plot for publication

I have a question regarding choosing only a certain number of features for the circus plot to make it more readable. We have used the sparse mode of DIABLO.

Here is a description of the data used for my Diablo model:
5 datasets:
$microbiome_genera 28 samples 143 features
$clinical_markers 28 samples 26 features
$feces_metabolome 28 samples1051 features
$plasma_markers 28 samples 1128 features
$plasma_ferulic_markers 28 samples 2 features

and selected 2 components for the final model with the following selection for each dataset and each component:

$microbiome_genera 4 1
$clinical_markers 3 15
$feces_metabolome 51 51
$plasma_markers 51 51
$plasma_ferulic_markers 1 1

cimDIABLO is working, here is the plot:

Code used for cimDIABLO:
color= hcl.colors(20, palette = “viridis”),
margins =c(15,30)#,

The network plot is working as well.

However, I am not able to select just component one in any of the function and How do we select top 10 features in component 1 of feces_metabolome perhaps?

We are asking becoz, currently, the circus plot is too big with too many features. For the publication, we would like to keep important features only.

hi @Aakash

You can have a look at ?circosPlot where you can specify the component you want to plot. And you can play with the cutoff threshold (however, that will apply for the strength of the association across all variables).

Another trick would be to rerun the DIABLO with the top selected features only and represent this. However, just be warned that you should only use this for representation purposes, not to report the performance as it will be too optimistic because you will be overfitting.


1 Like