Predictable difficulty or difficulty to predict

Abstract

Predicting the functional consequence of a given amino acid replacement, that is, the difference between the wild type protein and its nonsynonymous single nucleotide polymorph variants (nsSNPs) is a great challenge. This is especially true in the case of large membrane proteins, like ABC transporters. Kelly et al. has published such an ambitious study in a recent issue of Protein Science focusing on nsSNP in structurally conserved segments within the nucleotide binding domains (NBDs) of ABC transporters supposed to be involved in interdomain communication.1 With the aid of structural rationalization the authors predicted the impact of 40 nsSNPs found in seven clinically important human ABC transporters using a bivalent scoring: disease or neutral. We have performed a deep search of the available literature to find published data on the 40 nsSNPs to delineate them with the predictions. We have incorporated only data from original peer-reviewed publications into our dataset, and our hits were checked in various gene-specific databases too [BioBase and ABCC6 Database]. Based on the available information, we categorized the published data on a given nsSNP as (i) “disease-associated” when the publication clearly demonstrates that the variant segregates with the disease but is absent from at least 200 alleles of unrelated, nonaffected individuals; or (ii) in vitro experimental evidence when the specific variant was heterologously expressed and its (transport) function and/or folding-stability was investigated (Table I). We considered functional differences only if the level of expression of the given nsSNP variant was comparable to the wild type expressed in the same system. Our results, in comparison with the predictions of Kelly et al. are summarized in Table I. The left five columns of the table is taken from the publication of Kelly et al. On the right we list our findings and give the related primary reference. In several cases more than one independent studies were found showing the same result. To keep the list of references short, we indicated only one reference, if more publications were found we added +n (where n is the number of additional independent publications). We found the published data on 19 of the 40 ABC nsSNPs. Of these, 16 nsSNPs had been tested in in vitro experiments and in seven examples the given nsSNP resulted in impaired function and/or folding. In one case the published nsSNP generates a new splice site (exon skipping), what affects the mRNA maturation therefore the function or folding of the protein with the given nsSNP can not be concluded (ABCB11, E1186K). One of the nsSNPs was characterized solely by its association to a genetic disease (ABCC6, A1291T associated to pseudoxanthoma elasticum). One additional nsSNP was found as homozygous nsSNP in several healthy individuals therefore it was categorized as neutral, on the basis of this human genetic data (ABCC6, R1268Q). There are two nsSNPs with both published genetic association and in vitro experimental characterization in our data set (ABCB11 V444A associated to intrahepatic cholestasis of pregnancy; ABCG2, Q141K associated to gout). Our conclusions on the nsSNPs are listed in column “Published phenotype” of Table I.; for the sake of simplicity we kept the bivalent terminology of Kelly et al.: disease or neutral. Comparing the predicted phenotype of the Kelly's paper with the phenotype extracted from the published data reveals misprediction in roughly half of the cases (10/19) thus demonstrating the inherent difficulties of rationalizing and predicting the functional impact of snSNPs. This work was supported by the Hungarian research grants OTKA CK80135, OTKA NK81204, OTKA PD79783 and by NIH R01AR055225 (subaward) to A.V. A.T. is a recipient of Bolyai Fellowship of the Hungarian Academy of Sciences. Tamás Arányi*, Krisztina Fülöp*, Orsolya Symmons*, Viola Pomozi*, András Váradi [email protected]*, * Institute of Enzymology, Hungarian Academy of Sciences, Budapest, Hungary.