In other words, does the universe of discourse limit the interpretation of the predicate? So for example, say the universe of discourse is $\mathbb{Z}^+$ (positive integers). Let $P(x)$ be "$\sin(x) > 0$". If $x = 1$ then $\sin(x)$ is no longer in $\mathbb{Z}^+$. Does that mean $P(x)$ is false for those values of $x$ even though the inequality is true if you allow reals in the calculation? Or does the predicate exist independent of the universe of discourse and that universe applies only to the variables?
NOTE: if the universe of discourse was not stated explicitly or if we had a statement like $(\exists x\in\mathbb{Z}^+ \;|\; \sin(x) > 0)$ then I would not be concerned about the predicate "leaving" $\mathbb{Z}^+$.
PS. I've looked through all the "Similar Questions" that come up in the sidebar, I've searched for a bunch of variations on "predicate universe of discourse", and of course lots of Google searching, but can't seem to find a conclusive answer. I started getting in to ZFC axioms to try to sort this out, but that wasn't going anywhere either.
The question is: how are you defining/axiomatizing $\sin$? It is unlikely you can fully characterize $\sin$ in a theory where $\mathbb Z^+$ is a valid universe.
Ignoring that, let's say you just have a function symbol that you are calling $\sin$ and which you want to interpret as the function $\sin$. Then, if the universe was $\mathbb Z^+$, it would need to be a function $\mathbb Z^+ \to \mathbb Z^+$, but the only integer $\sin$ takes to an integer is $0$. Therefore this would simply not be a valid interpretation. (I'm also assuming $\mathbb Z^+$ contains $0$. If not, then you just end up with a totally undefined "function" and an empty relation below.)
You could instead use a predicate formulation such as $\sin(x,y)$ to mean intuitively mean $\sin(x)=y$, then you would not have an issue. This binary predicate symbol would be interpreted as a relation containing only $(0,0)$. With this predicate symbol, you could reformulate your formula as $\exists x.\forall y.\sin(x,y)\land y>0$. This would be false in the interpretation I suggested since $\sin(x,y)$ is only true for $x=0$ and $y=0$.