Molecular Database Systems/ Chemical Structure Search (AURAmol)
Introduction
Cybula has developed a powerful range of tools based on the
AURA high performance pattern recognition
technology for searching databases containing 3 Dimensional views
of complex small molecules (in the order of 60 atoms). The AURAmol
technology allows a user to search for molecules that have a similar
shape to a particular description. The basic technology underpinning
this task can be used in a wide variety of problems. Extensions to
the basic method allow the system to take into account local properties
of the molecules. More details of the system are now available on
the on-line demonstration web site.
Screen shot
The image below shows an typical example screen shot from a
demonstration front end desk top system. 200 molecules have been
taught from the NCI open database. A query has 'netropsin' has
been entered and 17 potential matches found in 0.32 seconds (on a
800MHz PC) as displayed in the list. Two example matches have been
displayed, along with the query (click on the screen shot for a detailed
view).
Principal features of the technology are:
- The technology allows very large databases of
molecules (> 100,000) to be searched.
- Input of new molecules to the database is
quick.
- The methods can be run on desktop workstations to
supercomputers matching the needs of the user.
- The technology uses the full 3D structure and
surface properties of the molecules
- The methods used within the technology are published,
allowing full understanding of the methods
The technology consists of a set of C++ functions built on top
of the AURA CMM library used in many of our systems. Typical use
is through UNIX/linux based programs. A simple JAVA based
demonstration system is available to show the operation of the
system.
Outline function
The AURAmol system describes the surface of the molecule by a
set of points. These points are joined in a graph that is then
used to search the database of molecules. The nodes in the graph
contain attributes that describe the local properties of the
molecule at that point. The match engine is composed of a number of
CMMs working together through a constraint propagation process.
The constraint update procedure has been developed specifically to
support CMM based systems, and efficiently searches large databases
for potential matches. The results of the process are then supplied
to the user with a measure of the similarity to each molecule
returned.
Technical details
The AURAmol system has been developed at the University of
York, Computer Science Department in association with GSK and
Evotec. The technology is described in detail in
a number of papers at the Advanced Advanced Computer Architectures
web site.
Use of the technology
The AURAmol system may be embedded into many applications as
well as run as a web service. The
system exists as a C++ library and runs on Linux, NT, Windows. The system can be run on small PCs to
supercomputers. The technology has been licensed to Cybula Ltd. and
is now available for incorporation into your systems.
|
HOME 
