Saccharomyces cerevisiae is the first eukaryotic organism for which a multi-compartment genome-scale metabolic model was constructed. Since then a sequence of improved metabolic reconstructions for yeast has been introduced. These metabolic models have been extensively used to elucidate the organizational principles of yeast metabolism and drive yeast strain engineering strategies for targeted overproductions. They have also served as a starting point and a benchmark for the reconstruction of genome-scale metabolic models for other eukaryotic organisms. A typical metric of validating genome-scale models is to by compare in silico growth predictions across different mutants utilizing different carbon sources with in vivo growth data. This in silico/in vivo comparison results in two types of model-prediction inconsistencies; either the model predicts growth when no growth is observed in the experiment (GNG inconsistencies) or the model predicts no growth when the experiment reveals growth (NGG inconsistencies). Further, comparison of the in silico growth predictions in presence of higher order gene deletions (i.e., synthetic lethals) with available experimental data reveals a number of additional ways that model and experiment may differ in their predictions. For example, GSL inconsistencies refer to cases where the simultaneous deletion of a gene pair results in a viable strain (i.e., Growth), where their deletion is lethal (i.e., Synthetic Lethal) based on experimental data. ESSL represent mismatches where the single deletion of one of the genes in silico is lethal (i.e., ESsential), however, their simultaneous deletion in vivo results in a lethal phenotype (i.e., Synthetic Lethal). Finally, SLG and SLES denote inconsistencies where the model implies that only the double gene mutation is lethal (i.e., Synthetic Lethal) but experimental observations support either growth (G) or lethality of any of the two single gene deletions (i.e., ESsential), respectively. In spite of the successive improvements in the details of the described metabolic processes, the yeast reconstructions were less predictive in these metrics when compared to contemporary E. coli reconstructions.
In this work we make use of the automated generalized GrowMatch procedure for restoring consistency with single gene deletion experiments in yeast and extend the procedure to make use of synthetic lethality data using the genome-scale model iMM904 as a basis. We identified and vetted using literature sources 120 distinct model modifications including various regulatory constraints for minimal and YP media. The incorporation of the suggested modifications led to a substantial increase in the fraction of correctly predicted lethal knockouts (i.e., specificity) from 38.84% (87 out of 224) to 53.57% (120 out of 224) for the minimal medium and from 24.73% (45 out of 182) to 40.11% (73 out of 182) for the YP medium. Synthetic lethality predictions improved from 12.03% (16 out of 133) to 23.31% (31 out of 133) for the minimal medium and from 6.96% (8 out of 115) to 13.04% (15 out of 115) for the YP medium. The resulting iAZ900 model contains 900 genes, 1240 metabolites and 1602 reactions distributed among seven compartments.
Zomorrodi, A.R. and C.D. Maranas (2010), "Improving the iMM904 S. cerevisiae metabolic model using essentiality and synthetic lethality data" BMC Systems Biology, 4(1): 178. PMID: 21190580.
The model is licensed under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Before downloading our model, we ask for your name and institution so we may better design future versions of our software. Every consideration has been made during the design, writing, and posting of this model to ensure it works properly.