How to automatically select a subset of attributes which names have been varied

Dear all,

I am about to reduce manual steps in my current process. But now I came to a point where I got stuck.

The original example set looks like this:
att1 att2 att3
1 3 6
2 4 7
3 5 8
4 6 9

After the first steps of processing (I do windowing) I receive a new set with more attributes:
label att1-0 att1-1 att2-0 att2-1 att3-0 att3-1

Then I want to split into "subsets" and this is where I struggle. Each subset shall include the label and all variations of each original attribute:
First iteration label, att1-0 and att1-1
Second iteration label, att2-0 and att2-1
Third iteration label, att3-0 and att3-1

My intention is to run a selection algorithm with each subset next and finally merge the remaining attributes in a single set again:
e.g. label, att1-0, att2-1, att3-0 (attributes according selection algorithm)


I already tried operators like "loop attributes" and  "work on subset" as well as combinations of both. But in the end I was not able to make it so far. Could anyone please provide a clue which approach could be promising?


Kind regards
Sachs

PS: The names and quantity of the attributes may change every time I run the process...