Use of decomposed data for Diablo


First of all, thank you very much for all your work and the useful tools that you developed.

I have been working with a three-level analysis (two fish strains, two conditions and three time points) using metabolomics and proteomics data. I used the DIABLO approach with some interesting results although with some difficulties to interpret as it seems there is an interaction.
On my journey, I found some tools to analyze each dataset separately using amOPLS in R. I was wondering if using the decomposed data off each main factor(strain, condition or interaction) could be used to feed DIABLO. In this way, my data will be “adjusted” to cover only the variability of each factor and that will facilitate the understanding of the correlation between datasets at each factor independently.

Thank you very much in advance!

hi @odrangel,

Yes, you can do this. The PLS methods that we have in the package do not allow currently to include any covariates, as you have correctly guessed, so sometimes you will need to process your data beforehand.
That said, you could include the strain / condition information in the Outcome Y in DIABLO to primarily focused on one of the effect, while ‘ignoring’ the other (if both strain and condition as not totally confounded).


1 Like