Figure 1. Purchasable chemical space is >>
1000-fold more bio-like than possible chemical
space. The GDB is a calculated library (Fink &
Reymond, 2005) of the 27 Million possible stable
molecules composed of C, N, O, F and H atoms, up to 11
atoms in size. The purchasable GDB is the subset that
can be purchased. Both sets are compared to a.
metabolites and b. natural products. From
Figure 2. Left A ~5 mM AmpC inhibitor
found by an NMR fragment screen. Right A 0.2 mM
AmpC inhibitor found by fragment docking (yellow
x-ray, green docking). The TINS fragment bound in a
site unsampled by docking.
Comparing the potential size of chemical space to
the actual size of screening libraries (Hann, 2001), it is a
miracle that screening ever works. One reason why it does,
we have argued (Hert, 2009; Figure 1), is that the libraries
that are actually screened are fortuitously biased towards
bio-relevant chemotypes. The chemical content of our
libraries, either for high-throughput, fragment,
DNA-encoded, or docking screens, is as important for the
success of a discovery campaign as is the quality of our
assays and the quality of our scoring functions.
An active project in the lab is to understand how to
enrich libraries with bio-relevant molecules, and with such
molecules grow the libraries. We have investigated how
docking compares to high-throughput screens (HTS) against
targets such as PTP-1B,
and how docking fragments
compares to docking larger, lead-like molecules, again using
as model systems. A very recent direction involves efforts
to grow our docking libraries by log orders. Already, at 3
to 6 million molecules, docking screens are larger than most
HTS campaigns; over the next several years we expect the
docking libraries to expand by 1000-fold, introducing
fascinating new challenges and huge opportunities.
A close coupling between theory, docking, and experimental testing is central to the enterprise. Key papers include:
Supported by NIGMS GM59957 & GM71896.