Specific interactions between carbohydrates (also known as glycans) and proteins underlie the initiation or
progression of many diseases. Carbohydrate-binding proteins (human, bacterial or viral lectins and adhesins)
and carbohydrate-processing enzymes (glycosyltransferases and glycosidases) are therefore important targets
for therapeutic intervention, however the creation of drug-like molecules that can competitively inhibit
carbohydrate-binding sites is uniquely challenging. The optimization of a glycomimetic inhibitor involves the
synthesis and screening of chemical analogs in an attempt to increase the inhibitory potential and biological
activity. Given that carbohydrate synthesis is notoriously laborious, the task of evaluating innumerable analogs
with incrementally increasing affinities introduces a particularly significant bottleneck for glycomimetic
development. Despite the challenges, the benefit of employing the native carbohydrate as a scaffold is that it
intrinsically confers the desired specificity. The fundamental challenge in the creation of a glycomimetic is that
of divining which modifications will lead to enhanced affinity without compromising specificity.
Computational approaches that are specifically designed to screen analogs of carbohydrates could be
invaluable aids to both increasing the objectivity of the synthetic choices and to prioritizing the synthetic effort
required for glycomimetic development. Virtual screening is commonplace in mainstream medicinal chemistry
and has led to the discovery of non-glycomimetic small molecule inhibitors with nanomolar affinities (12,29).
However, it has yet to be widely applied in glycomimetic design. We believe that this is due to several factors,
including the complexity of carbohydrate structure and nomenclature, which creates a significant barrier for
non-glycoscientists, and, for glycoscientists, a lack of familiarity with sophisticated modeling methods.
In the present application, we propose to develop, validate, and implement an alternative strategy to ligand
docking that leverages the benefits of computational modeling and structural biology. Specifically, we will
develop an online computational approach that uses carbohydrate-protein co-crystal (or NMR) structures as
the basis for lead optimization by modifying the bound oligosaccharide in situ. We have assembled a group of
experimental glycobiologists and chemists who have agreed to provide data and independently validate the
predictive accuracy of the tools we are developing. These scientists have over 200 years of combined
experience in glycomimetic synthesis and evaluation.
Successful completion of the aims will lead to a validated computational tool to aid in the discovery and
optimization of therapeutic agents that target carbohydrate-protein interactions that are particularly relevant in
the ongoing battle against multidrug resistant bacteria.