tsvali
commented
2 months ago Hi Mojit and thank you for your interest in RolDE! Indeed, adjusting the data with regards to the confounding effects by using suitable modelling approaches are potentially a valid option in dealing with such effects. Then, for a continuous covariate, after fitting the model for protein expression ~ covariate, the residuals from the model could be extracted as you suggested and used for further analysis with RolDE. In practice, there are many ways to do this and linear (mixed) models are definitely one good option. Of course, it is very much recommended to explore and visualize the results from the adjusted data with respect to the original data, the tested conditions/experimental groups and with respect to the adjusted confounding effects, and make sure the results seem reasonable. I hope this helps!
Best, Tommi
Hej!
We want to use RolDE on longitudnal proteomics data from Olink assays. The data is confounded by a continuous covariate as well as potential other factors. The instruction manual for RolDE says "By bare minimum, the user should provide RolDE the data in a normalized numerical matrix, adjusted for confounding effects if needed". I was wondering if you had recommendations regarding how to adjust for confounders beforehand?
For instance, would it be appropriate to model ~Covariate in a linear model/limma and extract residuals from the fit object and use it as input into RolDE?
Best, Mohit