Question about sample () in 2 group bootstrap test

My boss may want to know whether there is some significant difference between 2 groups (group1 and group2), and confirm the difference is not caused by thousands of gene numbers. So we plan to do a bootstrap test.

I saw some examples using sample() function with replace=TRUE. My question is that during the bootstrap is it reasonable to pool group1 and group2 together and reassigned them into group1 and group2 separately? Or should we pick group1's sample from group1 only?

For example,

I may have group1=c(1,2,3,4) and group2=c(7,8,9,10)

Is it possible/reasonable to have group1=c(1,1,7,9) and group2=c(3,4,7,7)?

Thank you very much