Applying Semantic Web Services to
bioinformatics: Experiences gained, lessons
learnt
Phillip Lord1, Sean Bechhofer 1, Mark D. Wilkinson2, Gary Schiltz3,
Damian Gessler3, Duncan Hull1, Carole Goble1, and Lincoln Stein4
1Department of Computer Science 2University of British Columbia
University of Manchester, James Hogg iCAPTURE Centre
Oxford Road St. Paul’s Hospital, 1081 Burrard St.
Manchester, M13 9PL Vancouver, BC, V6Z 1Y3
UK Canada
3National Center for Genome Resources 4Cold Spring Harbor Laboratory
2935 Rodeo Park Drive East 1 Bungtown Road
Santa Fe, NM 87505 Cold Spring Harbor, NY 11724
US US
Abstract. We have seen an increasing amount of interest in the ap-
plication of Semantic Web technologies to Web services. The aim is to
support automated discovery and composition of the services allowing
seamless and transparent interoperability. In this paper we discuss three
projects that are applying such technologies to bioinformatics: myGrid,
MOBY-Services and Semantic-MOBY. Through an examination of the
differences and similarities between the solutions produced, we highlight
some of the practical difficulties in developing Semantic Web services
and suggest that the experiences with these projects have implications
for the development of Semantic Web services as a whole.
1 Introduction
In the past 10 years, the ability to perform biological in silico experiments has
increased massively, largely due to the advent of high-throughput technologies
that have enabled the industrialisation of data gathering.
There are two principal problems facing biological scientists in their desire
to perform experiments with these data. The first of these is distribution–many
of the data sets have been generated by individual groups around the world,
and they control their data sets in an autonomous fashion. Secondly, biology is
a highly heterogeneous field. There are large numbers of data types and of tools
operating on these data types. Integration of these tools is difficult but vital. [2]
Biology has coped with this in an effective and yet very ad hoc manner. Al-
most all of the databases and tools of bioinformatics have been made available on
the Web; the browser becoming an essential tool of the experimental biologist.
The reasons for this choice of technology are partly chance in that the growth

2in genomic technologies happened to occur contemporaneously with the growth
of the Web. But many of the key benefits of the Web are also important for bi-
ologists. Publishing is economically cheap, technically straightforward, innately
distributed, decentralised, and resilient to change. Accessing the Web is likewise
simple, requiring no knowledge of specific query languages but enabling “query
by navigation” [6].
While this has worked well in the past, it has obvious problems. Many bioin-
formatics analyses use fragile screen-scraping technologies to access data. Keep-
ing aware of the Web sites on offer is, in itself, a full-time and highly skilled
task, mostly because of the complexity of the domain. The application of Se-
mantic Web services to bioinformatics seems a sensible idea in that Web services
provide a programmatic interface and avoid screen-scraping [14], while semantic
descriptions enable their discovery and composition.
In this paper we describe three architectures, myGrid, MOBY-Services and
Semantic-MOBY, which have been designed to address these problems. All three
are aimed mainly at bioinformatics. All three are based on Web or Web-services
technologies and use an additional specification of their services to describe the
semantics of their operations. All three are high-profile projects in the domain
of bioinformatics and come from groups with previous track records of providing
solutions for problems of interoperability.
The three solutions are also different in many respects from each other and
from the “idealised” Semantic Web services architecture. In examining these
differences, we raise a set of key questions about the applicability of Semantic
Web services in practice and present our (partial) solutions for these difficulties.
2 A day in the life: bioinformatics as it is
Bioinformatics as a discipline has largely grown directly out of the molecular-
biology laboratories where it was born. In general, each lab investigated a small
region of biology and there are very few labs world-wide working on a single
problem. Many of these labs have made their own data available for use on
the Web. This data is often un- or semi-structured. Much of the data is com-
posed of DNA or protein sequences, but this has generally been accompanied by
large quantities of “annotation”–descriptions (generally in free-text form) of the
sources of the sequences, literature citations and the possible function(s) of the
molecules. In addition to this raw data, many different tools that operate on it
have been developed, most of them with restricted functionality and targeted at
performing highly specific tasks.
This situation is slowly changing, largely due to the appearance of large ser-
vice providers in the form of the genome-sequencing and -annotating centres.
These centres are now increasing their scopes and often provide many different
types of information. The primary service consumer still remains the small
laboratory. Much of the information remains openly accessible.
The primary “integration layer” so far has been expert biologists and bioin-
formaticians. Using their expert knowledge of the domain, they will navigate