Unexpected result in plotVar comparatively to the optimized sPLS I performed

Hi everyone,

My project is about predicting, with transcriptional data (X), a disease severity score (Y).
I performed an sPLS that I optimize with the perf function.
The final 25 genes I have in my 2 first components seem to well represent the severity score in plotindiv function.

But when I performed plotVar function, my variables don’t look strongly correlated to my severity score.

Is there something I don’t understand or I did not do well in my analyses ?

Thanks a lot for your help.

dear @alain_LU,

The plotVar coordinates represent the correlation between each original variable, and its associated component. It might be worthwhile to check first the correlation between the components associated to X and Y to see if they strongly agree (for example with a scatter plot). Also remember that the correct interpretation should be ‘a linear combination of the selected variables is correlated with the severity score’, not individual variables themselves.

Hope that helps,