Biomedical Engineering Reference
In-Depth Information
24.4.1
Shared Computing Services
Distributed computing has gone through many phases starting with remote
procedure calls (RPCs) [64] in the 1970s-1980s to the Common Object Request
Broker Architecture (CORBA) [65] in the 1990s. More recently, distributed
computing in the form of so-called Web services have emerged. Because these
services expose methods, like RPC and CORBA, but also data, we consider
Web services to be a generalization, allowing one to provide access to both
data (contained in databases) as well as algorithms.
There are many protocols on which Web services can be based. These
include Simple Object Access Protocol (SOAP) [66], Representational State
Transfer (REST) [29], and Extensible Messaging and Presence Protocol
(XMPP) [67]. Most Web service protocols are designed to work with multiple
types of transport layers [Hypertext Transfer Protocol (HTTP), User Datagram
Protocol (UDP), etc.], but the majority of Web service protocols currently in
use work over HTTP. This approach results in signifi cantly easier deployment
and access of services since HTTP is ubiquitous across the Internet. For a more
detailed review of Web service technologies the reader is referred to Curcin
et al. [68] and Fielding et al. [69].
In this section we provide a brief overview of cheminformatics Web services
and some use cases highlighting the collaborative potential of such Web ser-
vices. Indiana University has developed a number of cheminformatics Web
services [70] that provide access to core cheminformatics methods (fi nger-
prints, 2D depiction, and various molecular descriptors), statistical techniques
(using R [71] as the backend), and chemical database access methods. Most of
these services are implemented in Java using the Chemistry Development Kit
[72] and SOAP as the underlying protocol. The services have been used in a
variety of scenarios. For example, the fi ngerprint and statistical model services
were employed by Indiana University to develop models to predict the activity
of user-supplied compounds against the NCI60 cancer cell lines [70].
Interestingly, the fi nal predictive model was itself converted to a Web service
and thus accessible by any SOAP client, remote or local. Note that, while the
model service was also located at Indiana University, it could easily make use
of fi ngerprint services located at other sites, anywhere across the world.
Recently, these Web services have been forked and reimplemented as
REST-based services. While SOAP services can be “hidden” behind a REST
interface, the reimplementation avoids the extra complexity imposed by SOAP
by directly exposing the functionality via the REST interface. These services
can be found at http://rguha.net/rest. Currently, the services are hosted at mul-
tiple locations and include Drexel University in the United States and Uppsala
University in Sweden and are utilized by a number of independent applica-
tions. An example is the use of these services to enhance an Open Notebook
Science (ONS) project [73], as reviewed elsewhere (see Chapter 25). The ONS
Solubility Challenge is an ongoing project in which a number of research
groups have experimentally determined the solubility of a variety of solutes in
Search WWH ::
Custom Search