Reactivity axis

Here we describe the "clean" axis in the exported subsets of ZINC15.

Traditionally, ZINC standard subsets included: A, B, C, D, E.

ZINC "clean" subsets were just A, B, C, D.

The Clean Axis

Class	Nickname	Description	How computed
A	Anodyne	No flags of any kind set	no pattern matches whatsoever
B	Unnecessarily complex	The worst feature is the molecules are over decorated e.g. too many chlorides	complex, requires representative in A. This is really reserved for future use, and gives us the opportunity to split "super clean" into A (representatives) and B (versions of A)
C	Chromophore	Worst problem is a chromophore that could interfere with some asays	extended conjugated system, pattern_type = 5
D	Any PAINS	Any pains pattern not covered in E,F,G,H	pattern_type= 8 (PAINS)
E	mildly electrophilic or nucleophilic	aldhydes, thiols, michael acceptors, epoxides	pattern_type in (1,2)
F	strongly electrophilic or nucleophilic	alpha halo ketones, alkyl halides. Note includes cancer drugs.	pattern_type in (3,4)
G	unstable in buffer	typically reagents, but could be used for covalent binding. e.g. boronic acids	pattern_type = 10
H	nonsensical	idea from the PAINS papers	pattern_type = 14

other concepts mentioned, must be fit in: chelation, redox, covalent, amphiphilicity

poor derivatizability, optimizability

we never build protomers of H, G, F.

we need to classify pains by assumed mechanism

397 pains never hit any compound in zinc