The Floating 'I Think'
Working with the sentence fragment that the poster provides, we can place "I think" in at least four locations with reasonably natural-sounding results:
There is an expression [that] I think comes from ...
or:
There is an expression [that] comes, I think, from ...
or:
I think there is an expression [that] comes from ...
or:
There is, I think, an expression [that] comes from ...
In the first two cases above, the "I think" attaches to the derivation of the (unnamed) expression and conveys the meaning, "I think a particular expression—call it expression X—comes from source Y." In the second two cases above, it attaches to the existence of the expression and has the meaning, "I think a particular expression—call it expression X—from source Y exists."
The author is confident enough in the expression's existence not to bother qualifying that part of the sentence with "I think"; but the same is not true with regard to the derivation of the expression. The two options that put "I think" in position to hedge the appropriate assertion are "There is an expression [that] I think comes from ..." and "There is an expression [that] comes, I think, from ..."
But in those two cases, the placement of the floating "I think" doesn't alter the underlying assertion that "I think" qualifies—namely, "There is an expression [that] comes from ..." For that reason, I have trouble seeing why omitting that in the first case is optional:
There is an expression [that] I think comes from ...
but omitting it in the second case is an error:
There is an expression [that] comes, I think, from ...
Another Possible Destination
So far, I've neglected a fifth location where "I think" would fit in the poster's sentence fragment without sounding weird—namely, immediately before the actual or implied that. Here it is:
There is an expression, I think, [that] comes from ...
This case, like the third and fourth cases above, uses "I think" as a qualifier attached to the assertion that the expression exists, rather than to the (incomplete) assertion about the source of the expression. But except for the commas and the position of the included that, it's identical to the first case above—the one from which the poster wants to omit that. In this fifth instance, too, omitting that yields a problematic result:
There is an expression, I think, comes from ...
A Brief Descent Into Style
Though much of what I've said to this point may seem to be championing the idea of making that explicit in "There is... that..." constructions, I'm not a big fan of those constructions. It's all too easy to get caught up in the issue of whether to include or omit that in a sentence like
There is something I think [that] doesn't love a wall.
to the exclusion of asking whether the "I think" is useful. (Frost evidently didn't think it was; but being a poet he also took the unprosaic step of transposing "there is" and "something.") Or for that matter, to the exclusion of considering whether both "There is" and "that" could be dispensed with for the benefit of the whole:
Something doesn't love a wall.
Such trimming doesn't always help sentences, and sometimes it seriously hurts them. But sometimes it does them good—especially when the writer tends to apply lard liberally and thoughtlessly to every page. This is a matter of style, however, and in style there is no truth.