APR Sept/Oct 2023 - 64

»
FORMULATION AND DEVELOPMENT
»
to all stakeholders at all times, and the integration of discovery
results with a variety of software-based statistical, modeling, and
documentation tools.
Until quite recently, computer-aided drug discovery was viewed as just
another tool to assist in the selection of " hit " molecules, performing
calculations, tabulating and tracking results, and managing data
overload. As the components of cloud computing emerged and
coalesced through interoperability, information technology evolved
from a tool to an active participant in drug discovery. Or, as one
commenter has put it, modern drug discovery has evolved from
computer-aided to computer-enabled.5
The marriage of cloud computing and drug discovery is
possible through:
*
In-silico simulation of select aspects of drug discovery
experiments such as protein binding and solubility
*
Interoperability of computational processes with physicochemical
assays
* Quantum mechanics (QM)-based computational methods
* Artificial intelligence for confirming anticipated trends and for
uncovering previously unknown drug-target interactions
Benefits of Cloud Computing
Cloud computing in drug discovery arose from the need to access
and consolidate computational tools that have evolved since the
late 1980s, particularly given the huge compound libraries involved
and the desire to screen those compounds electronically. The cloud
is where big data resources-including compound libraries and
applications that once were housed exclusively at user sites-will
eventually reside. Cloud computing provides the opportunity to draw
from any combination of these resources.
Computational approaches are broadly classified as either structurebased
or ligand-based. Analogous to physical high-throughput
screening, structure-based techniques require knowledge of both
target and ligand or test molecule. These techniques include, for
example, ligand docking, homology modeling and molecular
dynamics.6
Ligand-based discovery is more empirical in that it predicts
a test molecule's activity (both " efficacy " -related attributes and
toxicity) to molecules known to be active. These techniques include,
for example, pharmacophore detection and shape-based analog
searches. By serving as a repository for both data and applications, the
cloud empowers both approaches.7
The fundamental rationale for computational discovery methods
is the large number-millions or billions-of library entries and the
nearly infinite number and types of interactions they are likely to
encounter with targets. When hosted in the cloud these resourceintensive
processes become available across time and space, and are
scheduled to maximize CPU and GPU utilization.
More significantly, and unlike siloed applications, cloud computing
allows users to access and utilize custom, configurable computing
64 |
| September/October 2023
resources as networked services on demand.8
As such it addresses two
problems with traditional drug discovery.
By accessing deeper computational methods, cloud applications help
improve discovery efficiency by selecting test compounds, whether
they exist physically or only within virtual libraries, with superior target
binding and drug-like properties. The advantages are not just speed
and operational efficiency, but the discovery of novel chemical space,
which is particularly useful for targets that have traditionally been
difficult to " drug. " 9
The significance of covering novel chemical space cannot be
overstated, as existing libraries (again, whether housed in thousands
of microplates or as billions of computer entries) have already been
extensively mined. For example, conventional discovery has produced
very few usable target-ligand combinations for G protein-coupled
receptors. In the cloud, combining homology modeling with molecular
dynamics provides structural data which, combined with cloudbased
quantum chemistry applications and advanced searching of
novel chemical space, can lead to promising candidate molecules for
hitherto undruggable targets.10
Advanced Tools
The lure of computational approaches for drug discovery is the
potential for more rapid or accurate understanding of physical
ligand-target interactions than might be achieved through actual
experimentation. Numerous inputs are required, including modeling
based on molecular characteristics (e.g. shape, local charge,
lipophilicity, steric factors) and ligand-target complementarity. Crystal
structures were once the only way to do this but crystallography is
expensive and not every ligand-receptor pair can be easily crystallized.
Computer-aided drug design (CADD)
reproduces these events in
software, and has been successfully deployed in discovery programs
for several highly successful drugs.11
Early-generation CADD worked its magic through representations
of ligands and substrates in more-or-less conventional terms using
parameters like molecular topography and hydrophobicity. Secondgeneration
approaches, referred to as molecular mechanical methods,
incorporate conformational potential energies and electrostatics into
their algorithms, thus adding another layer to interaction analysis.
The major benefit of cloud computing to drug discovery, then, is
ready access to a host of on-demand software resources that can
vastly improve the scale and accuracy of computer-aided drug design
techniques and which users could otherwise acquire only at great
expense and effort.
Among these advanced tools are artificial intelligence (AI), machine
learning, and advanced quantum computing. AI is defined as " a
system's ability to interpret external data correctly, to learn from such
data, and to use those learnings to achieve specific goals and tasks
through flexible adaptation. " 12
AI is an umbrella term encompassing
computational methods that simulate human intelligence, especially
pattern matching.
APR Sept/Oct 2023

Table of Contents for the Digital Edition of APR Sept/Oct 2023
