Map PAINS substructures according to PAINS class and find the highest PAINS violation

Many scientists are unaware that PAINS substructures are classified by hit
frequency; this has led to a tendency to irresponsibly use PAINS filters
during compound triage by filtering out potential leads based on the
presence of a PAINS substructure instead of the presence of a type of
PAINS substructure. Baell suggested in his review, Seven Year Itch:
Pan-Assay Interference Compounds (PAINS) in 2017 - Utility and
Limitations, that scientists should focus on class A, maybe class B, and
not class C. Because the original paper only reported hit frequency based
on a singular assay technology, compounds should not be necessarily
excluded for containing a class A substructure. This protocol splits the
different classes of PAINS into distinct substructure mapping subprotocols
to produce a more informed PAINS report. The SMARTS patterns can be
obtained here. This protocol is being made available in conjunction with
an upcoming publication.