Limits...
Comparative Analysis of Yeast Metabolic Network Models Highlights Progress, Opportunities for Metabolic Reconstruction.

Heavner BD, Price ND - PLoS Comput. Biol. (2015)

Bottom Line: We have also compared pairwise gene knockout essentiality predictions for 10 of these models.We found that varying approaches to model scope and annotation reflected the involvement of multiple research groups in model development; that single-gene essentiality predictions were affected by simulated medium, objective function, and the reference list of essential genes; and that predictive ability for single-gene essentiality did not correlate well with predictive ability for our reference list of synthetic lethal gene interactions (R = 0.159).We conclude that the reconstruction of the yeast metabolic network is indeed gradually improving through the iterative process of model development, and there remains great opportunity for advancing our understanding of biology through continued efforts to reconstruct the full biochemical reaction network that constitutes yeast metabolism.

View Article: PubMed Central - PubMed

Affiliation: Institute for Systems Biology, Seattle, Washington, United States of America.

ABSTRACT
We have compared 12 genome-scale models of the Saccharomyces cerevisiae metabolic network published since 2003 to evaluate progress in reconstruction of the yeast metabolic network. We compared the genomic coverage, overlap of annotated metabolites, predictive ability for single gene essentiality with a selection of model parameters, and biomass production predictions in simulated nutrient-limited conditions. We have also compared pairwise gene knockout essentiality predictions for 10 of these models. We found that varying approaches to model scope and annotation reflected the involvement of multiple research groups in model development; that single-gene essentiality predictions were affected by simulated medium, objective function, and the reference list of essential genes; and that predictive ability for single-gene essentiality did not correlate well with predictive ability for our reference list of synthetic lethal gene interactions (R = 0.159). We conclude that the reconstruction of the yeast metabolic network is indeed gradually improving through the iterative process of model development, and there remains great opportunity for advancing our understanding of biology through continued efforts to reconstruct the full biochemical reaction network that constitutes yeast metabolism. Additionally, we suggest that there is opportunity for refining the process of deriving a metabolic model from a metabolic network reconstruction to facilitate mechanistic investigation and discovery. This comparative study lays the groundwork for developing improved tools and formalized methods to quantitatively assess metabolic network reconstructions independently of any particular model application, which will facilitate ongoing efforts to advance our understanding of the relationship between genotype and cellular phenotype.

Show MeSH

Related in: MedlinePlus

Change in gene essentiality predictions between model and its nearest ancestor.When comparing the Matthews Correlation Coefficient for model gene essentiality predictions to the models’ nearest progenitors, we observe that the models may be segregated between those focusing on expanding model scope, and those focused on iterative refining an existing model by plotting the change in MCC between models. Generally, when the stated focus of a model developer is to expand the scope of the yeast metabolic network reconstruction, predictive ability suffers relative to the progenitor model. When the stated focus is to refine and curate a model, predictive ability improves relative to the progenitor model. Thus, our analysis finds that model predictive ability reflects the iterative process of model development. The asterisk near the Yeast 4 comparison indicates that it is an integrative model that not have a single nearest progenitor (we compared it to iFF708 for this analysis).
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4643975&req=5

pcbi.1004530.g003: Change in gene essentiality predictions between model and its nearest ancestor.When comparing the Matthews Correlation Coefficient for model gene essentiality predictions to the models’ nearest progenitors, we observe that the models may be segregated between those focusing on expanding model scope, and those focused on iterative refining an existing model by plotting the change in MCC between models. Generally, when the stated focus of a model developer is to expand the scope of the yeast metabolic network reconstruction, predictive ability suffers relative to the progenitor model. When the stated focus is to refine and curate a model, predictive ability improves relative to the progenitor model. Thus, our analysis finds that model predictive ability reflects the iterative process of model development. The asterisk near the Yeast 4 comparison indicates that it is an integrative model that not have a single nearest progenitor (we compared it to iFF708 for this analysis).

Mentions: Improving model ability to predict the essentiality of individual genes for growth has not been the primary motivating factor for developing each new yeast metabolic network model, but essentiality predictions have generally been reported with the publication of each new model to demonstrate their accuracy and utility. However, direct comparison between reported predictive values is complicated by differing simulation conditions. In this study, we did not find a strong trend of continuing improvement in model ability to predict single gene essentiality over time. Instead, we found evidence of an iterative process in which model scope changes (typically leading to a decrease in average predictive accuracy), followed by subsequent curation leading to improved prediction by descendant models (Fig 3). We found that both iIN800, with its expansion of the reconstruction of lipid metabolism, and iND750, with its expansion of compartmentalization, had a lower overall Matthews Correlation Coefficient (MCC) for single-gene essentiality predictions than their progenitor model, iFF708. Subsequently, iMM904 refined iND750, and made more correct predictions of single gene essentiality. Similarly, Yeast 6 refines Yeast 5 and improves predictive ability, but the focus of Yeast 7 on expanded scope does not lead to as large an improvement in single-gene essentiality predictive ability, and iTO977’s focus on expanding model scope to cover some protein modification processes and to provide a scaffold for integrating transcriptomic data does not lead to an improvement in predicting single-gene essentiality compared to its progenitor model, iIN800.


Comparative Analysis of Yeast Metabolic Network Models Highlights Progress, Opportunities for Metabolic Reconstruction.

Heavner BD, Price ND - PLoS Comput. Biol. (2015)

Change in gene essentiality predictions between model and its nearest ancestor.When comparing the Matthews Correlation Coefficient for model gene essentiality predictions to the models’ nearest progenitors, we observe that the models may be segregated between those focusing on expanding model scope, and those focused on iterative refining an existing model by plotting the change in MCC between models. Generally, when the stated focus of a model developer is to expand the scope of the yeast metabolic network reconstruction, predictive ability suffers relative to the progenitor model. When the stated focus is to refine and curate a model, predictive ability improves relative to the progenitor model. Thus, our analysis finds that model predictive ability reflects the iterative process of model development. The asterisk near the Yeast 4 comparison indicates that it is an integrative model that not have a single nearest progenitor (we compared it to iFF708 for this analysis).
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4643975&req=5

pcbi.1004530.g003: Change in gene essentiality predictions between model and its nearest ancestor.When comparing the Matthews Correlation Coefficient for model gene essentiality predictions to the models’ nearest progenitors, we observe that the models may be segregated between those focusing on expanding model scope, and those focused on iterative refining an existing model by plotting the change in MCC between models. Generally, when the stated focus of a model developer is to expand the scope of the yeast metabolic network reconstruction, predictive ability suffers relative to the progenitor model. When the stated focus is to refine and curate a model, predictive ability improves relative to the progenitor model. Thus, our analysis finds that model predictive ability reflects the iterative process of model development. The asterisk near the Yeast 4 comparison indicates that it is an integrative model that not have a single nearest progenitor (we compared it to iFF708 for this analysis).
Mentions: Improving model ability to predict the essentiality of individual genes for growth has not been the primary motivating factor for developing each new yeast metabolic network model, but essentiality predictions have generally been reported with the publication of each new model to demonstrate their accuracy and utility. However, direct comparison between reported predictive values is complicated by differing simulation conditions. In this study, we did not find a strong trend of continuing improvement in model ability to predict single gene essentiality over time. Instead, we found evidence of an iterative process in which model scope changes (typically leading to a decrease in average predictive accuracy), followed by subsequent curation leading to improved prediction by descendant models (Fig 3). We found that both iIN800, with its expansion of the reconstruction of lipid metabolism, and iND750, with its expansion of compartmentalization, had a lower overall Matthews Correlation Coefficient (MCC) for single-gene essentiality predictions than their progenitor model, iFF708. Subsequently, iMM904 refined iND750, and made more correct predictions of single gene essentiality. Similarly, Yeast 6 refines Yeast 5 and improves predictive ability, but the focus of Yeast 7 on expanded scope does not lead to as large an improvement in single-gene essentiality predictive ability, and iTO977’s focus on expanding model scope to cover some protein modification processes and to provide a scaffold for integrating transcriptomic data does not lead to an improvement in predicting single-gene essentiality compared to its progenitor model, iIN800.

Bottom Line: We have also compared pairwise gene knockout essentiality predictions for 10 of these models.We found that varying approaches to model scope and annotation reflected the involvement of multiple research groups in model development; that single-gene essentiality predictions were affected by simulated medium, objective function, and the reference list of essential genes; and that predictive ability for single-gene essentiality did not correlate well with predictive ability for our reference list of synthetic lethal gene interactions (R = 0.159).We conclude that the reconstruction of the yeast metabolic network is indeed gradually improving through the iterative process of model development, and there remains great opportunity for advancing our understanding of biology through continued efforts to reconstruct the full biochemical reaction network that constitutes yeast metabolism.

View Article: PubMed Central - PubMed

Affiliation: Institute for Systems Biology, Seattle, Washington, United States of America.

ABSTRACT
We have compared 12 genome-scale models of the Saccharomyces cerevisiae metabolic network published since 2003 to evaluate progress in reconstruction of the yeast metabolic network. We compared the genomic coverage, overlap of annotated metabolites, predictive ability for single gene essentiality with a selection of model parameters, and biomass production predictions in simulated nutrient-limited conditions. We have also compared pairwise gene knockout essentiality predictions for 10 of these models. We found that varying approaches to model scope and annotation reflected the involvement of multiple research groups in model development; that single-gene essentiality predictions were affected by simulated medium, objective function, and the reference list of essential genes; and that predictive ability for single-gene essentiality did not correlate well with predictive ability for our reference list of synthetic lethal gene interactions (R = 0.159). We conclude that the reconstruction of the yeast metabolic network is indeed gradually improving through the iterative process of model development, and there remains great opportunity for advancing our understanding of biology through continued efforts to reconstruct the full biochemical reaction network that constitutes yeast metabolism. Additionally, we suggest that there is opportunity for refining the process of deriving a metabolic model from a metabolic network reconstruction to facilitate mechanistic investigation and discovery. This comparative study lays the groundwork for developing improved tools and formalized methods to quantitatively assess metabolic network reconstructions independently of any particular model application, which will facilitate ongoing efforts to advance our understanding of the relationship between genotype and cellular phenotype.

Show MeSH
Related in: MedlinePlus