Tomiuk, Stefan;Scheel, Hartmut;Hofmann, Kay - Exploiting weak sequence similarities for protein functional prediction

ECCB 2002 Poster sorted by: Author | Number

Next | Previous poster (in order of the view you have selected)

Title: Exploiting weak sequence similarities for protein functional prediction	P166
Tomiuk, Stefan; Scheel, Hartmut; Hofmann, Kay stefan.tomiuk@memorec.com Bioinformatics Group, MEMOREC Stoffel GmbH, Stöckheimer Weg 1, D-50829 Köln, Germany

Sophisticated techniques for sequence comparison, including the 'generalized profile' method or the 'profile HMM' method, are nowadays widely used to identify weak but statistically significant protein sequence similarities [1]. It is generally believed that the three-dimensional structures of evolutionarily related proteins are much better conserved than the respective sequences. While this assumption is certainly true in some cases, it sometimes appears to be a misconception based on the use of inappropriate sequence comparison methods and/or on the overestimation of visually perceived structural resemblance. Only recently, rigorous statistical criteria are being applied to structural comparisons, similar to the criteria that have been commonplace in sequence comparisons for a long time. Here, we present two examples that underscore the usefulness of sequence-based comparison in the elucidation of protein function.
The first example describes the sequence analysis of a recently published bacterial phospholipase A2, which is functionally similar to eukaryotic secreted PLA2 types and very likely shares a common ancestor with these proteins. A similarity of the corresponding sequences can readily be found by standard generalized profile methods and is associated with a high degree of statistical significance. Interestingly, the 3D-structure of the bacterial enzyme is fundamentally different from the structure of the eukaryotic PLA2 enzymes, where only the arrangement of the active site residues is conserved. Consequently, none of the structural comparison programs tested was able to find any similarity (significant or not) between the two structure classes, meaning that a functional assignment of the novel bacterial PLA2 would not have been possible from structural data alone. To our knowledge, this is one of the first examples of a detectable sequence similarity between functionally related proteins in the absence of detectable structural similarity.
The second example describes a novel homology domain [2] that we identified in a number of proteins, including two apparently unrelated gene products that - when mutated - give rise to very similar disease phenotypes belonging to the idiopathic epilepsies. The two proteins in question, MASS1 and LGI1, are the only known gene products associated with idiopathic epilepsy, which are not ion-channels. Our finding points towards an alternative disease mechanism that is likely to be associated with the novel repeat domain, which we call the 'EAR-domain'. Sequence analysis and structural predictions suggest that the EAR repeat is a member of the seven-bladed beta-propeller fold.

[1] Hofmann K. (2000) Sensitive Protein Comparisons with Profiles and HMMs. Briefings Bioinf. 1: 167-178 (2000).
[2] Scheel H, Tomiuk S, Hofmann K. (2002) A common protein interaction domain links two recently identified epilepsy genes, Hum. Mol. Genet. 11:1757-1762.