SDAP: database and computational tools for allergenic proteins

INTRODUCTION

Allergic diseases, including allergic rhinitis, asthma and atopic dermatitis, are among the most common chronic health problems (1). As recombinant proteins, products of the genomic revolution are introduced into foods, medications and other products of our daily life, distinguishing allergens from other proteins becomes a pressing issue (2). This need was recently illustrated by the concern over possible allergenic effects of Starlink corn (3). The sequences and structures of many allergenic proteins have been determined. Most of these proteins can be grouped into relatively few families (4,5), suggesting that they share characteristics that contribute to their ability to bind IgE and trigger an allergic response (6,7). The web server SDAP (Structural Database of Allergenic Proteins) was created to aid in identifying these sequence commonalities. Computational and statistical tools to analyse sequences and structures, assessable within SDAP, have been designed to develop correlations that can be used to predict allergenicity of novel proteins and cross-reactivity between allergens.

The computational approach is briefly depicted in Figure 1. The sequences, 3D structures and IgE epitopes of known allergens, collected from databases and literature, are included in lists in the server. For allergens where no experimental structure has been determined, models will be computed with our MASIA/EXDIS/DIAMOD/FANTOM suite of programs (8–17).

Using this database and the computational tools from SDAP we will develop predictive models for IgE epitopes. This will allow a user to compare not just sequence or property similarity of a possible epitope homologue, but also factors such as local structure and relative surface exposure. SDAP is available at http://fermi.utmb.edu/sdap/.

DATABASE STRUCTURE

SDAP is designed as a web server (Fig. 2) controlled by a set of CGI scripts. These scripts mediate interaction with the user, the database and the computational tools. The database component of SDAP is implemented with MySQL under Linux. The information is collected in tables according to: allergen type; species; systematic name; brief description; sequence accession numbers from SWISS-PROT/TrEMBL, PIR-ALN, NCBI Taxonomy Browser and, where available, PDB. Sequences and IgE epitopes are collected into text files. The current lists of allergens were assembled from literature and from major sequence [SWISS-PROT/TrEMBL (18), PIR-ALN (19) and NCBI Taxonomy Browser (20)] and structure [PDB (21)] databases, guided by the list of allergen names from the IUIS website, http://www.allergen.org. As there is no public database summarizing information on known epitopes of allergenic proteins, the IgE epitope list in SDAP is based solely on primary literature sources. In its present version, SDAP allows searches restricted to the following fields: allergen name (according to the IUIS website listing), scientific and common name for the species and general source of the allergens. SDAP is the first allergen database that allows a user to retrieve IgE epitopes and identify similar regions in other allergenic proteins.

COMPUTATIONAL TOOLS

Special programs have been incorporated in SDAP to compare the sequences and structures in the database. In the present release, users can compare a given peptide sequence to all the sequences in SDAP, using either an exact match search or a similarity search based on physicochemical properties (22). The SDAP peptide exact match function is useful to identify allergens closely related to a peptide, for example when an IgE epitope may be responsible for clinically-defined cross-sensitivities to several allergens.

One use of SDAP is to quickly identify cross-reactivities between known allergens. For example, an exact match search using the Pen i 1 IgE epitope MQQLENDLDQVQESLLK from shrimp topomyosin identified the same sequence in the allergens Met e 1 (from another shrimp species) and Pan s 1 (lobster), consistent with the clinically observed cross-reactivity among these crustaceans (see Table 1). However, sequence identity, even among know cross-reactive allergens, is rare. To identify more distantly related sequences, the user can access a tool that uses the amino acids descriptors E₁–E₅ (22) to locate sequences with similar chemical properties. The vectors were derived by multidimensional scaling of 237 physical–chemical properties for all 20 naturally occurring amino acids. The mathematical procedure used ensures that the main variations of all 237 properties for the 20 amino acids are reflected by the five descriptors E₁–E₅. Using the E₁–E₅ descriptors, the similarity between two sequences A and B, each one consisting of N residues, is computed with the property distance function PD (23):

where λ_j is the eigenvalue of the jth E component, E_j(A_i) is the E_j value for the amino acid in the ith position from sequence A, and E_j(B)_i is the E_j value for the amino acid in the ith position from sequence B.

The SDAP tool calculates the PD similarity index between the query sequence and each sequence-window with the same length from all allergens collected in the SDAP protein database. The search result is a list of similar sequences identified in allergenic proteins, presented in decreasing order of similarity (increasing PD) with the query sequence. Besides epitope identification, this tool can be used to find conserved regions in the allergens from the SDAP protein database.

The results of a sequence similarity search for the above Pen i 1 IgE epitope are shown in Table 1. Besides the two exact matches with Met e 1 and Pan s 1, similar regions were found in allergenic tropomyosins from insects (storage mite, American cockroach, American house dust mite and European house dust mite), lobster, crab, abalone and snail. Clinical tests with sera from sensitive patients indicate that the cross-reactivity between crustacean, mollusk and insect allergens is mediated by tropomyosin (24,25). Sensitization to house dust mites had been linked to oral allergic response to snails (7) and allergen immunotherapy with European house dust mite (Dermatophagoides pteronyssinus) may trigger severe reactions to mollusks and crustacea (26). These results demonstrate the utility of SDAP in identifying potential cross-reactivity among allergens.

The significance of any sequence match can be determined from a histogram of the distribution of PD values for all the sequences in SDAP to the shrimp epitope (Fig. 3). The sequences in Table 1 clearly have a lower PD value than the bulk of entries in SDAP. The histogram suggests that a significance cut-off value between 7.5 and 9 would be most appropriate for determining peptides with similar properties.

DATA SUBMISSION

Researchers in the allergy field are welcome to submit their published data by email to oiivanci@utmb.edu. Comments, corrections and suggestions for new computational tools for allergenic determinants should be sent to the same address.

FUTURE DEVELOPMENTS

The SDAP server will be maintained on a regular basis. The database sequence, structure and IgE epitopes lists will be updated with information as it becomes available in other databases and the literature. The next major addition to SDAP will be data and software to compare the structures of epitopes. The 3D structure is known for only about 10% of the sequences in SDAP. For the sequences without structures, homology models will be prepared with our self-correcting distance geometry based EXDIS/DIAMOD/FANTOM suite. To determine whether suitable templates were available, the sequences of 180 allergens in SDAP with unknown 3D structure were submitted to the fold recognition server 3D-PSSM (27) (http://www.sbg.bio.ac.uk/~3dpssm) and a histogram of the distribution of the log(E) values for the best template found in the PDB (Fig. 4). For 150 sequences, a good template [log(E)<−1] was identified. For the remaining allergens, eight have a log(E) between −1 and 0, meaning that the modeling may require combined information from other fold recognition servers. For these, we plan to use our MASIA program (8) (http://www.scsb.utmb.edu/masia/masia.html) to identify conserved motifs, which can be used to identify possible templates and improve alignments with distantly related proteins. Only 22 sequences, with log(E)>0, will require an alternative approach. Here, we will use secondary structure prediction methods to model the proteins abinitio.