Science requires methodological naturalism if it is going to use evidence to test (and perhaps falsify) hypotheses. The evidence must be observable and, as others have pointed out, repeatable.
Such evidence must be "natural" mustn't it? I'd like an explanation of how it can not be. From this methodological naturalism follows.
The testing against evidence that science uses to discern wrong from possible right requires methodological naturalism. Science doesn't exclude a transcendent designer specifically. The designer is excluded, as other point out, by the very people who want to suggest it as an explanation for anything.