words)A transcriptional feedback loop is central to clock function in animals, plants and fungi. The clock genes involved in its regulation are specific to -and highly conserved within -the kingdoms of life. However, other shared clock mechanisms, such as phosphorylation, are mediated by proteins found broadly among living organisms, performing functions in many cellular sub-systems. Use of homology to directly infer involvement/association with the clock mechanism in new, developing model systems, is therefore of limited use. Here we describe the approach PREMONition, PREdicting Molecular Networks, that uses functional relationships to predict molecular circadian clock associations. PREMONition is based on the incorporation of proteins encoded by known clock genes (when available), rhythmically expressed clockcontrolled genes and non-rhythmically expressed but interacting genes into a cohesive network.After tuning PREMONition on the networks derived for human, fly and fungal circadian clocks, we deployed the approach to predict a molecular clock network for Saccharomyces cerevisiae, for which there are no readily-identifiable clock gene homologs. The predicted network was validated using gene expression data and a growth assay for sensitivity to light, a zeitgeber of circadian clocks of most organisms. PREMONition may be used to identify candidate clockregulated processes and thus candidate clock genes in other organisms.Keywords: circadian clock / interactome / S. cerevisiae / network / circadian rhythm the absence of identified clock genes. For instance, Neurospora crassa shows free-running rhythms and circadian entrainment without the clock gene frequency 3,4 and rhythms in the redox state of peroxiredoxin persist in many model genetic systems in the absence of transcription or of clock genes 5 . Paracrine signaling can sustain circadian rhythms in pacemaker (neuronal) tissue that lacks clock genes and displays no endogenous rhythm 6 . One interpretation of this is that feedback integral to circadian clocks spans multiple molecular 'levels', from transcription to metabolism. We conjecture that this multi-level response can be captured through functional relationships, referred to as functional interactions. We know of no models that explicitly attempt to integrate these different levels to more fully describe the mechanisms of the circadian clock. Here, we describe the method PREMONition, PREdicting MOlecular Networks, to construct multi-level (functional) molecular clock networks using publicly available datasets and bioinformatics tools. After tuning the method on clock genetic model systems, we deployed PREMONition to predict a molecular clock network in S. cerevisiae, an experimental system with a circadian phenotype that is not suited to genetic screens 7 . We validated the results of the yeast experiment in silico and in vivo. Methods like PREMONition, that span multiple levels in the cell, are essential to developing integrative models of complex biological processes.
Results
PREMONition applied to mod...