Information Technology Reference
In-Depth Information
Allergenicity prediction has been attempted by several groups. The methods are
explained and discussed in the next section.
5.3.1 Sequence Similarity Searches
Sequence similarity search methods are an obvious approach to predicting
allergenicity. If two proteins are highly similar and one of the proteins is an allergen,
the likelihood of the second protein being an allergen is high. Sequence similarity
search methods are very mature and easy to implement. However, allergenicity is
determined by the binding of epitopes. Since epitopes are in general subsequences of
the entire protein sequence, local alignments like BLAST and FASTA (Pearson
1994) tend to perform better. In fact, the FAO/WHO guidelines implement local
alignment methods. Sequence similarity search methods are implemented in SDAP
and FARRP as means to query the content.
Even for identifying cross-reacting allergens, local alignment methods are useful
because cross-reacting allergens are generally more than 70% identical in their
sequences (Aalberse 2000). With a few exceptions, cross-reacting allergens share
similar protein structures (Aalberse and van Ree 1996). The main drawback of
sequence similarity searches is that their performance is limited to linear epitopes.
Conformational epitopes, unlike linear epitopes, are not composed of consecutive
amino acids. Therefore, sequence similarity search methods are not applicable to
conformational epitopes. The similarity in 3D structure often translates to similar
primary protein sequences.
Another drawback of sequence similarity search is its dependence on the
coverage of the dataset the query sequence is searched against. This makes the
detection of novel allergens difficult and requires comprehensive and constantly
updated allergen databases. Nevertheless, sequence similarity search methods have
also been implemented as a last-resort method for the prediction of allergenicity
(Stadler and Stadler 2003; Li, Issac, and Krishnan 2004).
5.3.2 FAO/WHO Guidelines
The FAO/WHO guidelines for allergenicity predictions involve a bioinformatics
component. This serves as an initial screening process prior to the use of any laboratory
test. Compared to laboratory testing, allergenicity prediction using bioinformatics
approaches is comparatively fast and simple. Implementation of the guidelines is
relatively simple and the computational complexity is low. However, the method due to
its reliance on primary protein sequences similarity is only as good as the underlying
dataset used. This means that allergen databases that contain the allergen dataset
are in a good position to implement these guidelines as a service. In practice, only
SDAP and Allermatch (Fiers, Kleter, Nijland, Peijnenburg, Nap, and van Ham 2004)
implemented the guidelines.
The FAO/WHO guidelines produce results that tend to have very high recall
and very low precision (Hileman, Silvanovich, Goodman, Rice, Holleschak,
Astwood, and Hefle 2002; Kleter and Peijnenburg 2002; Stadler and Stadler 2003;
Soeria-Atmadja, Zorzet, Gustafsson, and Hammerling 2004). Consequently, the
