PAINS questioned

Peter Kenny @pwk2013 just brought an interesting paper to my attention:

Privileged Scaffolds or Promiscuous Binders: A Comparative Study on Rhodanines and Related Heterocycles in Medicinal Chemistry

In short, hydantoins, rhodanines, thiohydantoins, and thiazolidinediones might not be as problematic (PAINs-y) as is often widely assumed. Interesting data, definitely worth a look before you throw out that baby with your bathwater.

This supports our design of ZINC15 where we offer four levels of "cleanliness" in screening libraries (and many more if you use the API). The levels are, in increasing order of stringency:

* Reactive - often stable under assay conditions, these tend to react easily and non-specifically, although it does include some drugs, e.g. alkyl halides are cancer drugs (DNA alkylators)

* Mild - the default, was called "standard" in ZINC12. This includes compounds such as alkyl thiols and michael acceptors, which often do not react under assay conditions, and can be both good leads and drugs.

* Clean - removes the thiols, michael acceptors, nitriles and some other compounds containing weak nucleophiles, electrophiles or compounds with mild redox activity. Notably, this level allows (includes) compounds with most PAINS patterns.

* Anodyne - ok, we're being a bit provocative with this name. This level removes every compound containing any PAINS pattern, giving you a "safe", "boring" and narrow screening library. Caveat emptor!

You will meet two kinds of people. One group thinks we should all be screening with Anodyne libraries, where everything that anyone could object to has been removed: "saves time, avoids artifacts" is their refrain. The second recognize that chemistry and biochemistry are complex, that there are exceptions to every rule, and that when you are screening you want to 1) get hits and 2) do controls. These people will tend to prefer the clean, mild and reactive subsets, depending on the project.