How to interpret the weights explained across each omics block under each component in DIABLO object?

Hi,

I have been exploring DIABLO for integrating 4 omics types using different design matrices (weighted and full). In DIABLO objects of my analysis, I see the below. What does the $weights mean here? Is there any reference material that describes how to interpret these weights?

diabloFulldes$weights
               comp1     comp2
mRNA       0.7840607 0.5215328
proteinO  0.6266970 0.5001603
proteinS  0.5434681 0.4893959
metabolite 0.5143882 0.3633174
> diabloWtdes1$weights
               comp1     comp2
mRNA       0.3545887 0.5443691
proteinO  0.4175279 0.5101271
proteinS  0.4444788 0.5067080
metabolite 0.3408058 0.3480089
> diabloWtdes2$weights
               comp1     comp2
mRNA       0.7840607 0.5215328
proteinO  0.6266970 0.5001603
proteinS  0.5434681 0.4893959
metabolite 0.5143882 0.3633174

It says from the help that it is the Correlation between the variate of each block and the variate of the outcome. Used to weight predictions. So can this be interpreted as mRNA analytes (selected) in comp1 and comp2 correlate with phenotypic outcome 78.4% and 52.15% in diabloFulldes while in diabloWtdes1 the correlation of mRNA analytes in comp1 and comp2 with the phenotypic outcome is 35.4% and 54.4% respectively?

hi @vd4mmind,
Apologies for the delay. It means that the component has a correlation of 0.78 with the categorical outcome, i.e. it is quite good to discriminate your sample groups. It is usual to observe a better discrimination with a full model, compared to a weighted model (we discuss this in the DIABLO paper).

Thank you for using mixOmics DIABLO!

Kim-Anh

1 Like