The short version is it’s doable, but it takes some doing to accomplish. The example in that link snaps world position to match a pixelated UV rather than the other way around, which you could try to reverse engineer. The main issue is UVs are in an arbitrary space unrelated to world space, and it’s not entirely trivial to convert from one to the other, which is what the above linked thread is attempting.
The other option is to not use the UVs at all, but rather use a world position and xy size, which would already be in world space, to calculate the alpha from. For a sprite, you’d have to have a c# script that sets this on the material as you move it around. But if you can use a regular mesh instead you can extract this information from the mesh’s transforms along with knowledge about how big the original mesh’s bounds are to ensure you don’t try to make a circle larger than is visible. But then you loose a lot of the various UI related features which you may want or need.