Chemistry Reference
In-Depth Information
sets to augment Merck's internal sample collection. An important part of the library design
process is the evaluation of the chemical properties of the available fragments or reagents
and their impact on the fully synthesized library. In addition to chemical property distri-
butions, novelty and diversity may be considerations when spanning chemical space or
developing molecules to elucidate structure-activity relationships (SARs).
With the goal of making library design tools widely available to medicinal chem-
ists throughout Merck's Research Laboratories, a web-based library design platform was
developed. The web-based method allows users from anywhere on the Merck intranet
to access the latest versions of the code without downloading and installing specialized
software and allows the developers to restructure and optimize code in response to user
feedback and system load.
These web-based library design tools are collectively known as the Virtual Library Tool
Kit (VLTK), comprising four major components: (i) Reagent Selector (RS) (ii) Synthon
Analysis (SA), (iii) Library Enumeration (LibE) and (iv) LibraryAnalysis (LA). These have
been discussed in detail elsewhere.
[
41 43
]
Considerable amounts of research and resources
were devoted to developing appropriate functional queries and selecting an optimal set of
reagents, resulting in tools which distinguish RS and SA from earlier efforts. Seemingly
trivial queries such as 'find all primary amines' are found to yield different results when
using MDL's ISIS
[
44
]
and Daylight's SMARTS.
[
45, 46
]
A comprehensive discussion of the
salient points in forming typical medicinal chemistry queries is available.
[
41, 42
]
A pictorial
methodology for selecting reagents based on synthon properties was also developed to facil-
itate fragment selection.
[
42, 43
]
Synthons are usually defined as the fragments of the reagents
remaining in the final products. However, for our purposes it is also important to define a
point of reference for determining how the synthon will impact the properties of the final
products. This point of referencewill be the atom(s) of the synthon participating in the newly
formed bond(s). In the virtual fragment scan for primary amines in the second BACE-1
example above, the primary amine of the fragment is the reference atom. By combining our
unique visual approach for synthon properties with synthon structural clustering, one may
quickly select a set of fragments with desired properties, e.g. diversity, molecular weight,
number of hydrogen bond acceptors and donors, etc., from the universe of potential candid-
ates. An additional complication arises for multi-dimensional libraries where the properties
of the final compounds result from two or more synthetic groups. In this case, the universe
of each synthetic group may be chosen so that the final distributions of properties are within
acceptable limits of the user-specified constraints. This is accomplished in VLTK through
the implementation of the GLARE algorithm recently published by Truchon and Bayly.
[
47
]
The library design tools simplify the process of searching the available reagent data-
bases for fragments, selecting the most interesting subset, performing the virtual library
enumeration and evaluating the final virtual library properties. If the library also has a spe-
cific target, as in the present example, the VLTK tools also assist conformation generation,
docking and scoring.
Fragment selection is accomplished with the RS component of the VLTK. While the
selection of fragments can be a time-consuming process, prone to error and missteps, RS
allows the user quickly to select fragments with a 'point and click' approach. Intrinsic
filtering tools provide the user with fragments sets that have a high likelihood of being
available at a reasonable cost in the quantity and purity needed. This implementation of
VLTK's user-directed fragment selector is unique in its focus on the structure and properties
