Bioinformatics and biochemistry
Posted by sspiro on October 2, 2007
The unexpected appearance of this blog in Bio::Blogs 15 (the bioinformatics blog journal) prompts me to ramble some more on the topic of bioinformatics, and specifically gene annotation and re-annotation, from the perspective of an ‘experimental’ micro/molecular biologist.
I moaned recently about a published study in which obviously incorrect conclusions about gene function were reached on the basis of naïve sequence comparisons. Pawel made some suggestions as to how (computationally) mistakes like this might be avoided. In the particular case I was bothered about, some common sense application of biological understanding should also have prevented the error. This was a case not of incorrect annotation during a genome sequencing project, but rather incorrect re-annotation by investigators looking at single genes on a case-by-case basis.
As Pawel mentioned, we should expect and tolerate annotation errors in large-scale sequencing projects. An interesting case is described in a recent paper in the Journal of Bacteriology from James Ferry’s lab. Here, a gene annotated as encoding a carboxymuconolactone decarboxylase is instead shown to encode a enzyme (MdrA) that has protein disulfide reductase activity and contains an [Fe-S] cluster. Correction of all annotation errors is not a realistic goal if it requires painstaking biochemistry of the type described in this paper. But doing some biochemistry will improve the accuracy of computational predictions; the two approaches are (obviously) synergistic. Interestingly enough, from the way that the paper is written it seems that Ferry’s study was stimulated by the annotation error. The presence of the gene in a cluster of others probably related to oxidative stress was one clue to the annotation error (physiology of the organism was another). So, if the gene had been elsewhere in the genome, and incorrectly annotated, then this work may never have been done!
There is some very interesting biochemistry in the paper as well. The authors show that the disulfide reductase activity and oligomeric state of MdrA are regulated by an iron-sulfur cluster. A Cys-X-X-Cys motif is required both for cluster formation and for disulfide reductase activity, but the [Fe-S] form of the protein is apparently catalytically inactive. The data lead the authors to hypothesize that MdrA trimers are crosslinked to form an enzymatically inactive hexamer by a single [2Fe-2S] cluster. In this model, loss of the cluster (perhaps in response to oxidative stress) leads to the formation of trimeric MdrA with disulfide reductase activity.
