This is low priopryy but we could generate BioGRID annotatiosn for the GO MF with substrate data,
and use the BioGRID biochemical activity interactions to inform out MF_> has_direct input annotations.
Need to think about this more but wanted to remember so opening a ticket now