¿ìè¶ÌÊÓƵ

IT IS the year 2050, and you are flicking through a booklet of advertisements
for the latest gadgets, looking for a present for the friend who has everything.
You spot it; a compact disc holding their complete genome – a personal record
of the DNA that helps to mould their physical and mental characteristics.

Such a present might well be feasible by then, although for most people
it would have value only as a novelty item. Even for the scientist, the
genome of one individual would not be ideal. What medical science needs
is a generalised genome, the genetic equivalent of the generalised skeleton
used to depict our anatomy, based on NA instead of bone.

The human genome project is a Pounds sterling 2 billion venture involving
scientists from around the globe. Stretching over the next 15 years at least,
the project aims to produce just such a generalised genome, from which geneticists
will try to establish the differences between individuals and ‘the norm’.
Biochemists will use the sequence to plan novel genetically engineered organisms
that can produce scarce, medically useful human proteins and hormones. Doctors
and researchers will use the clues in our DNA to discover, diagnose, understand
and treat human genetic diseases.

But making sense of this information will be a complicated task, partly
because of the sheer quantity of data the project will churn out. The genome
comprises 3500 million molecular groups known as ‘bases’, or nucleotides.
These are known as A (adenine), C (cytosine), T (thymine) and G (guanine).
When arranged in sequences of tens of thousands of bases, maybe more, these
nucleotides represent genes. These genes are arranged on chromosomes, of
which each individual has 46 arranged in pairs. There are 24 different chromosomes,
including the two sex chromosomes.

The first five years of the project will be devoted to producing a physical
map of the human genome, showing only its major landmarks. At the end of
1989, molecular biologists agreed to store their information for the map
in the same way, regardless of how they obtain their results. They adopted
a system of labelling known as sequence tagged sites, which is a form of
genetic shorthand. The system reduces the amount of information that needs
to be stored on computer and it allows different groups of researchers on
the genome project to know what each other has done or is doing. This approach
has been made possible through the polymerase chain reaction, or PCR, a
technique in which fragments of DNA are cloned in a test tube (‘Genes unlimited’,
¿ìè¶ÌÊÓƵ, 14 April 1990).

As the project progresses, molecular biologists will produce more details
about the sequences of DNA in the genome – the way the strings of four bases
are laid down in a specific order that dictates the function of each gene.
One of the computer scientists involved in the project is Chris Rawlings
of the Imperial Cancer Research Fund (ICRF) in London. He imagines the sequence
of bases as a huge compost heap of leaves that must be sorted, and stored
logically on computers, perhaps according to their position on our chromosomes.
‘Our job is to find a way of putting all of the leaves in the heap back
onto the tree . . . and in the right order,’ says Rawlings.

The current collection of DNA sequences represents only 1.5 per cent
of the human genome, but this is only part of the information that must
be stored. The database will grow between 200 and 300 times by the year
2000 as scientists submit newly decoded sequences plus supporting evidence
and additional notes, such as initial scientific interpretations, references
to the literature and even photographic evidence of experimental results.

The total amount of data is not trivial, but neither is it overwhelming.
CDs can already hold about 500 million bytes of information and so, with
each base using a byte of storage space, the entire human genome would fit
onto just six or seven of them. By the time the genome is unravelled, CDs
will probably be advanced enough to carry the whole thing on a single disc.

By far the biggest challenge is to ensure that computer scientists gather,
collate and store the haphazard submissions of information from around the
globe in a coherent way. But computing for the biological sciences is a
very young discipline and those involved in the genome project have yet
to decide on the services best suited to their needs. ‘The computing aspects
of the project are of enormous importance,’ says Lennart Philipson, director
general of the European Molecular Biology Laboratory (EMBL) in Heidelberg,
West Germany. ‘If we can’t store the information and make it available then
you could question the need for a human genome project at all.’ The EMBL
runs the Nucleotide Sequence Data Library, one of the world’s two main collecting
points for genetic information. The other main database, GenBank, is in
the US; the National Institutes of Health currently uses the Los Alamos
National Laboratory, an agency of the US Department of Energy, to collate
information for GenBank.

The EMBL sends out its data on CD-ROMs, which are updated quarterly.
It uses the worldwide scientific data network, BITNET, and a newly established
European Molecular Biology network, EMBnet. The laboratory’s database also
includes references to other databases, such as the Protein Identification
Resource in Washington DC and the Protein Data Bank of 3-D structures of
proteins at the Brookhaven National Laboratory in New York. GenBank and
the EMBL database also exchange information with the DNA Database of Japan
at the National Institute of Genetics in Mishima. The DDBJ is funded by
Japan’s Science and Technology Agency.

One of the difficulties with designing databases for the genome project,
and setting their operating rules, is that the project’s requirements are
always changing. This sets them apart from commercial databases, such as
stores of invoices or credit notes, which are classified according to a
simple coded reference number, or name, that does not alter. On the genome
project, improvements to the machines that decode our genes will allow scientists
to submit data at an accelerating rate, often before they have had a chance
to make even a preliminary guess at what the data means. A fresh scientific
interpretation will put new demands on the databases. Researchers will spot
analogies between genes in different organisms, which they will want to
store. They will understand more about the function of polymorphisms – the
genetic differences between individuals – and will want to represent these
on computer. Also, scientists working on different aspects of the genome
will make different demands on the available information. Some will want
data on viruses alone, or on coding sequences with specific functions, such
as those that code for a particular disease. The result is that those managing
the databases will have to add, refine, correct and supply knowledge about
the sequences in a variety of ways that changes continuously.

Advances in the way that scientists use databases should help the next
Human Gene Mapping workshop, which the ICRF will host in London in 1991,
to be more productive. These biennial workshops bring together nearly a
thousand of the world’s leading biologists. Since 1973, when the first workshop
was staged, delegates have spent most of the time at these workshops shuffling
reams of written results among themselves instead of discussing what the
results mean. The next workshop will be different. For almost a year before
the workshop begins, the biologists will be able to use their own computer
workstations to record the fruits of two years’ research onto a database
at Johns Hopkins University in Baltimore. This will leave them free at the
workshop for discussion, debate and possibly some scientific agreement.

A network of databases

The Genome Database at Baltimore, which is connected to two academic
networks covering Europe and the US, operates on a smaller scale than those
that store DNA sequences; its main purpose is to act as a reference library
for the information that is available at these other databases. The GDB
is coordinated by the university’s William Welch Library and supersedes
the Human Gene Mapping Library, a database of outdated design .

The ICRF is preparing for next year’s workshop by bringing together
around 70 biologists for a preliminary session at the University of Oxford
in September. This group will help to ensure that their colleagues feed
information into the Baltimore database as systematically as possible. Some
of the group will represent committees responsible for each of the 24 different
pairs of human chromosomes. Some will dictate topics such as nomenclature,
and draw up instructions on how to name the DNA sequences. Others will devise
methods of storing information about spurious fragments of DNA that occupy
important sites on a chromosome but whose exact sequence is unclear. The
group will ensure that links are ready to be established as new databases
are created to contain information on the genomes of other organisms, such
as the mouse and nematode worm. These could be useful in creating comparative
models for human diseases.

Johns Hopkins University is also the birthplace of another important
database for gene mappers, the Online Mendelian Inheritance in Man, which
holds clinical data on a range of genetic diseases. OMIM is a computerised
version of a classic book of molecular biology, Mendelian Inheritance in
Man by Victor McKusick. The genome database at Baltimore is now cross-referenced
to OMIM. McKusick, a professor at the university, also instigated the formation
of the Human Genome Organisation, which was set up in 1988 to help to coordinate
the various national human genome projects.

The US is more advanced than Europe with its development of databases
for the genome. Each of the American databases has a specific function.
Some hold only sequences of DNA; others record the sequences of amino acids
that make up our proteins, or they store the 3-D structures of these proteins.
Others contain electronic indexes to the reams of published literature that
surround the project.

American scientists have prepared their politicians to expect to spend
around $200 million a year on the project for at least the next 15 years.
Two government organisations, the Department of Energy and the National
Institutes of Health (NIH), will channel this money to researchers. Tension
between the two over which one of them should administer the databases of
DNA sequences indicates how valuable this information is to the project.
But the rivalry has subsided following the emergence of the NIH’s National
Library of Medicine in Washington DC as the leading contender for the management
of databases in a structured, coordinated way – as the project demands.
A group of mathematicians and biologists at the library takes over the GenBank
database from the Los Alamos National Laboratory in 1992.

To the molecular biologist, databases of genetic information are merely
research tools. The trouble is that using them will become more difficult
as their number increases in an uncoordinated fashion around the world,
which is what is happening now. Jim Ostell, head of software technology
for the group at the National Library of Medicine, says that its first task
was to set up a fresh database, GENINFO. This brings order to the disparate
collections of information on the genome, which can change from day to day.
According to Ostell, GENINFO will act as a static ‘database of record’,
a permanent backbone of information to which other, more specialised databases,
including GenBank and the one at Heidelberg, can refer. The library will
label each piece of information with the date on which it was saved, and
leave it intact. Ostell estimates that GENINFO is growing at a rate of 60
000 to 100 000 sequences, of varying length, every year. GenBank contains
only around 35 000 sequences.

The challenge of this scale of librarianship is immense. GENINFO must
be able to accept information from a diverse range of sources without changing
its basic structure so often that software companies cannot keep up with
writing programs that can quickly recover particular types of information.
Software specialists at the National Library of Medicine, working under
David Lipman, produce their own programs to pick through the database. In
its latest test, Lipman’s software took just 10 seconds to find a sequence
5000 bases long. A standard search, which would look through the entire
database, would take several hours. Other groups, notably computer scientists
at the University of Wisconsin and at the EMBL in Heidelberg, have produced
similar software. They are developing programs that will make quicker and
more specific searches of genetic databases.

Biologists on the project will also require new computer hardware, and
this is likely to include specialised chips working together as parallel
processors. ‘The way to do this is to get molecular biologists working in
close proximity to computer scientists. Until now, computer science has
been devoted mainly to problems of physics or chemistry. Here is the chance
for it to turn to biology,’ says Philipson.

Teams working on the EMBL database, GenBank and the DNA Database of
Japan already work closely together. In February, they ironed out differences
in the way they represent information in their databases. Although, previously,
they had not duplicated the task of recording new data, corrections or revisions
were often made at all three sites. In future, the three databases will
be revised automatically, which will save time.

This agreement, however, still leaves much work to be done. Around 70
per cent of sequences sent to the EMBL come directly from scientists, complete
with their own notes in a standard style, and mostly in a format that allows
computers to read the information into the database. But a team of postgraduates
must still laboriously pick out other information on these sequences from
scientific journals – and annotating this information with supplementary
notes is the most labour-intensive task at the library in Heidelberg. ‘We
have had tremendous problems and disagreements over overlapping regions
of DNA,’ says Graham Cameron, who runs the library. ‘Joining together fairly
distantly related sequences may be appropriate for one researcher’s needs
but entirely unacceptable to another.’

One worrying development for the scientists at the EMBL is the way that
biologists will soon be able to think of the two main stores of information,
in Washington and Heidelberg, as a single database. The Heidelberg team
is anxious that this does not allow the US to gain a monopoly on gathering
and disseminating the information. It vigorously defends the need for an
independent European database. Philipson is campaigning for a European Institute
of Bioinformatics along the lines of the European Space Agency, or CERN.
Such an institute, he suggests, should be funded by the European Community
and sited close to the EMBL. ‘Everybody realises it is necessary, but no
one will provide the funds,’ he says.

The scientific community is only just beginning to tackle questions
of the intellectual property rights to the information in the project’s
databases. Some researchers argue that scientific journals should publish
only the work of scientists who also submit their sequence data to the databases.
Philipson argues that the project relies on databases being able to disseminate
new information quickly. He is concerned about the lengthy patenting processes
in both Europe and the US that he claims can hold up publication by several
months.

One recent development may help. Intelligenetics, an American software
company already involved in the operation of GenBank, is collating a new
database, GENESEQ, which will bring together sequences contained in applications
for patents in more than 30 countries. The database will provide information
long before it is published and much earlier than either GenBank or the
European database in Heidelberg. The first version of this database should
be available from the end of the year. Intelligenetics says that the information
in GENESEQ will complement that held in GenBank and at the EMBL, which often
includes details of industrial applications rather than just the bald information
contained in patents.

Philipson is optimistic that the genome project will provide the pressure
needed to prompt a revolution in the patenting of decoded genetic sequences,
and perhaps of patenting in general. He says that biologists must be able
to obtain patents within a week so that others can benefit from their work.

Developments in databases are impossible to predict, given the pace
of change in the capabilities of computer hardware and software. Cameron
says that the successful databases of the future will be fundamentally different.
They will distinguish between the database of scientific findings and the
database of genetic information stored by the cells or organisms themselves.
He argues that these two concepts are confused in many of the existing databases.
‘This will be a new kind of research with its own triumphs, errors, ambiguities
and developments. The scientific findings are the backbone, and must be
stored securely. Their interpretation will be the subject of endless debate.’

* * *

Time to consider a change of relations

RELATIONAL databases, such as the Genome Database at Baltimore, first
emerged as commercial systems in the late 1970s. They store information
in two-dimensional tables, and allow scientists the flexibility to view
their data grouped together in a variety of different ways. Such databases
include internal encyclopaedias that dictate the nature of the relations
between the data held in their tables.

Relational databases are currently the most popular form of database
on the genome project, but some computer scientists are looking at alternatives.
The main contender is the ‘object-oriented’ database, which many scientists
already agree is a potentially more powerful tool.

With an object-oriented database, programmers can store data as an abstract
object – the database does not force them to convert the information into
facts and figures that will fit neatly into a flat table.

An object can comprise many other objects, but the database will retrieve
all of them as a whole. The database can easily store abstract relationships
between these objects, and can cluster similar objects together. This would
be useful if a biologist wanted to search a database for a particular class
of genetic codes.

Eventually, the data produced by the genome project will appear in a
mixture of forms, including strings of bases, tabular references and even
photographs of experimental results. It is more difficult to accommodate
such variety in a relational than in an object-oriented database. As a result,
many of the computer groups that are searching for ways to store genetic
data are dabbling in object-oriented programming as a possible solution.

A team in Britain at the Medical Research Council’s Laboratory of Molecular
Biology in Cambridge is trying to unravel the 100 million bases that make
up the genome of the nematode worm. The group has decided to use an object-oriented
approach to store this information, and is developing its own customised
database using a popular object-oriented programming language, C++.

Elsewhere, computer scientists are adopting still more esoteric techniques
for constructing and searching genetic data bases. Japanese researchers
are tackling the genome project with the fruits of their effort to develop
a fifth-generation computer – one that emulates the way humans understand
information by processing knowledge rather than numbers.

Japan’s fifth generation computing project is very far from achieving
its goals, but it has produced a successful research computer, the multi-PSI,
and the software to exploit the machine. The computer consists of many processors
operating simultaneously, in parallel. It is particularly suited to sifting
through large databases to compare the information these hold with fresh
data. This is exactly what molecular biologists will demand of databases
supporting the genome project.

The head of the Japanese programme, Shunichi Uchida, says his researchers
have already tested their approach by installing roughly a quarter of the
data held at GenBank onto their database management system, Kappa. They
now hope to extend the use of their computers on the genome project through
a collaboration with the Argonne National Laboratory in the US. The Japanese
Institute of New Generation Computer Technology will install terminals at
Argonne, through which researchers will communicate with a multi-PSI in
Tokyo.

Ross Overbeek, who heads the team at Argonne, is a member of the Joint
Informatics Task Force in the US that oversees the development of information
systems for the genome project. He is championing the potential of parallel
processors using the so-called ‘logic programming’ techniques pioneered
in Japan. Overbeek says, along with other groups including computer scientists
at the EMBL, that this is the best way to tackle the welter of information
that the genome project will generate.