trans

14 PARASITE GENOMICS
loss of many of the most critical phenotypes
of the parasite as experienced in the wild.
In addition, much research is carried out
on model parasites, often natural parasites of
model laboratory hosts. A conflict can then
arise between the desire to know the genotype
of the disease organism, and the utility of
knowing the genotype of the well tested model.
Parasite genome projects have struggled with
this conflict and different outcomes have
been based on balancing competing needs. For
filarial nematodes, the choice of human parasite
was simple: only one can readily be maintained
in laboratory culture, and one ‘strain’
of this species is almost universally used in
research: thus the TRS strain of B. malayi was
chosen. For leishmaniasis, the many species
causing different disease syndromes suggested
that it might be necessary to examine multiple
genomes, and even multiple strains within
each: a compromise was reached once it
became evident that Leishmania species share
extensive synteny, and thus a well studied
L. major strain was chosen. Users of genome
data should be aware of this necessary simplification
in considering and using genome
data: the genotype determined may not reflect
the highly variable genotypes of real-world
populations.
BIOINFORMATICS AND THE
ANALYSIS OF GENOME
DATASETS
The completion of a genome sequence is not
the final product. The wealth of data encoded
in the millions of contiguous bases must be
interpreted and linked to biology. The size of
genome datasets has required the development
of a new way of analysing data in biology, generally
grouped under the term bioinformatics.
The linear sequence of the genome DNA
encodes (in response to the environment) the
four-dimensional organism. Bioinformatics
aims to ‘compute’ the organism given the DNA
sequence.
All of the parasite genome projects include
a dedicated bioinformatic component, and
learning the concepts and skills of bioinformatics
is essential if parasitologists are to
exploit the data in their research. The informatics
of interpretation of linear strings of
data, in terms of pattern recognition and
‘emergent’ higher order properties, has been
well developed outside biology, but biologydriven
informatics is now a dynamic and fruitful
field. The genome sequence can be likened
to a book, where each character has been
painstakingly determined, but where we have
little idea of the language, syntax and grammar
in which it was written. Bioinformatics aims to
derive language, grammar and syntax from the
data. The project starts from some universals,
such as the genetic code, and a general view of
structural features such as exons and introns,
but the rest has to be derived from the data. For
example, given a conflicting set of exon predictions
(open reading frame segments flanked by
valid splice sites), is it possible to predict with
accuracy the correct gene model, and thus the
correct encoded protein sequence? Questions
such as these are yielding to ever improving
gene-finding and predictive algorithms. Features
such as overall sequence complexity, the
presence of local or global repeats, the pattern
of di- or tri-nucleotide sequences, the relationship
between repeats and predicted exons, and
the pattern of exonic versus intronic or extragenic
DNA can now be computed and used to
annotate a genome.
Many of the tools used are based on probabilistic
methods, and thus need a training set
of ‘known’ genes from the organism of interest
in order to start extracting information. As
a genome project advances, the training set
MOLECULAR BIOLOGY