Genome-scale functional prediction and hypothesis testing in S.
cerevisiae highlight rRNA biogenesis as a major under-studied
process.
Lani Wu, Nevan Krogan, Armaity Davierwala, WenTao Peng, Mark Robinson,
Jack Greenblatt, Steve Altschuler, Timothy Hughes
BBDMR, University of Toronto, 112 College St., Toronto, ON M4T 2J4,
Canada
Six years after genome sequencing revealed 6,000 protein-coding yeast
genes, over 2,000 remain largely uncharacterized. To systematically
generate hypotheses regarding the cellular roles of these remaining
proteins, we developed a new method for making functional predictions
that utilizes publicly-available microarray data and categorical
annotation databases. The performance of the method in each functional
category was calibrated by examining false-positive and false-negative
prediction rates among 3,400 proteins with established cellular role.
Among poorly-characterized yeast proteins, we make 2,368 predictions
(encompassing 1,644 proteins) anticipated to be between 30% and 71%
accurate. To test our predictions, we are initially focusing on
ribosomal RNA processing, which is one of the largest prediction
categories (285 poorly-characterized proteins, comprising ~5% of all
yeast proteins) and is also the one with the largest proportion of
conserved and essential proteins. We have created a microarray to
monitor processing of rRNA and other non-coding RNAs, and we are
creating TET-promoter alleles of all of the essential genes.
Furthermore, we are TAP-tagging all of these proteins for MS analysis of
purified complexes, which are also being analyzed on the microarray for
RNA content. As of April 1, >50 proteins have been analyzed, revealing a
variety of rRNA processing defects. Hence, one of the most fundamental
cellular processes is also one of the least well-studied.
Return to YGM 2002 Home at SGD