A Maestro at Work

See allHide authors and affiliations

Science  14 Apr 2006:
Vol. 312, Issue 5771, pp. 163
DOI: 10.1126/science.312.5771.163c

In global surveys of proteins, from those based on sequence to those based on function, mitochondria have often lost out, in part because of the small proportion (7%) of cellular proteins that localize to this organelle. Calvo et al. set out to remedy this gap in proteomics by integrating their analysis over eight data sets, each of which is organized along a different dimension: mitochondrial targeting sequence, protein domain, transcriptional regulatory element, yeast homology, similarity to Rickettsia (the nearest living relative), coexpression, mass spectrometry, and proliferation induction. These data were used to train a Bayesian classifier, the Maestro, that when challenged with the Ensembl set of 33,860 human proteins, properly predicted 71% of the known mitochondrial proteins.

On a smaller scale, Maestro was applied to a human mitochondrial disorder—hepatic mitochondrial DNA depletion, in which the loss of mitochondrial DNA leads to organ failure—that had been mapped to a region on chromosome 2 containing 150 annotated genes. Spinazzola et al. sequenced the highest scoring candidates and found one, MPV17, for which mutations segregated with affected individuals in three unrelated families. They show that the absence of this inner mitochondrial membrane protein results in deficits in mitochondrial DNA and oxidative phosphorylation in mice. — GJC

Nat. Genet. 38, 10.1038/ng1776; 10.1038/ng1765 (2006).

Navigate This Article