How?
The way you do this is … by not doing it. There’s not really any good way to offset a UV space position to map to a 3D space or 2D view / screen space position because they don’t align in any sane way. Mesh UVs by their nature are arbitrary. The usual solution to something like is to use some kind of projection and ignore the mesh UVs entirely. Most commonly done by using projected screen space coordinates. Or by using a 2D projection based on the object’s local space and picking some arbitrary “surface” position within that space. In the later case you’ll need to do a raycast from c# to find what position on the surface of the mesh the cursor is aiming at and feed the back into the shader via material / global shader properties.