Continuous variable prediction from a MultiOmics dataset?

Question

[Aurelie MLB]

Hello,

Would anyone know a good package/published method to predict a continuous variable from a multiOmics dataset please? The possibility to understand the features driving the prediction is really important. I have found DIABLO. But it seems to only be suitable to the prediction of a categorical variables.

Many thanks Aurelie

[Jean-Karim Heriche]

It looks like you want to build a regression model. This can be done in a number of ways, for example using random forest like in this paper. However if you're interested in the features, as an alternative, you could look into factor analysis. Factor analysis of multi-omics data relies in essence on tensor factorization for which there are different approaches and formulations. For a Bayesian approach, check for example, the MOFA Bioconductor package or it's new version on GitHub.