Information Technology Reference
In-Depth Information
8.3 SERVICEMAP: PROVIDINGMAP AND
GPS ASSISTING SERVICE COMPOSITION
IN BIOINFORMATICS
The wide use of Web services and scientific workflows has enabled
bioinformaticians to reuse experimental resources and streamline data
processing in a Web-scale manner. This section presents a follow-up
work of the network analysis on myExperiment as discussed in the
previous one. The motivation comes from two common questions
proposed by scientists: (1) Given the services that I plan to use, what
are other services usually used together with them? and (2) Given two
or more services I plan to use together, can I find an operation chain to
connect them based on others' past usage? Aiming to provide a
system-level GPS-like support to answer the two questions, we
present ServiceMap [224], a network model established to study
the best practice of service use. We propose two approaches over
the ServiceMap: association rule mining and relation-aware, cross-
workflow searching. Our approaches are validated by the real-life data
obtained from the myExperiment repository. Empirical statistics of
the constructed service network are also reported.
8.3.1 Motivation
Our experience in the caBIG project has revealed two questions that
domain scientists frequently ask when they try to exploit external Web
services in building a scientific workflow:
Q1
: Given the services I plan to use, what are the other services
often used together with them, by other scientists?
Q2
: Given two or more services I want to use together, can I find
an operation chain, which is already used by others, to connect
them?
Q1 is usually raised when scientists obtain some data or analytical
capabilities wrapped and exposed as services. Since they are new to the
services, they intend to know how their peer scientists use them together
with other services of which they may or may not be aware. Besides,
due to the explorative nature of scientific workflows, incorporating a
