Since the release of ModelSEED in 2010, the systems biology research community has used the ModelSEED genome-scale metabolic model reconstruction pipeline to build over 200,000 draft metabolic reconstructions that support hundreds of publications. Here we describe the first comprehensive update to this reconstruction tool, with new features such as (i) a dramatically improved representation of energy metabolism, which ensures that models produce accurate amounts of ATP per mol of nutrient consumed; (ii) a new template for Archaea model reconstruction; and (iii) a significantly improved curation of all metabolic pathways with mappings to RAST subsystems annotations. We applied the improved pipeline to build and analyze core and genome-scale models for Archaea and Bacteria genomes in KEGG. The new ModelSEED pipeline generates larger models that possess more reactions and genes and require fewer gap-filled reactions. In addition, we see conserved patterns in the ATP biosynthesis mechanism across phylogeny, and identify clades where our understanding of energy biosynthesis is still poor. The ModelSEED v2 pipeline is currently available only as new reconstruction and gap-filling Apps in the KBase platform.

For over 10 years, ModelSEED has been a primary resource for the construction of draft genome-scale metabolic models based on annotated microbial or plant genomes. Now being released, the biochemistry database serves as the foundation of biochemical data underlying ModelSEED and KBase. The biochemistry database embodies several properties that, taken together, distinguish it from other published biochemistry resources by: (i) including compartmentalization, transport reactions, charged molecules and proton balancing on reactions; (ii) being extensible by the user community, with all data stored in GitHub; and (iii) design as a biochemical ‘Rosetta Stone’ to facilitate comparison and integration of annotations from many different tools and databases. The database was constructed by combining chemical data from many resources, applying standard transformations, identifying redundancies and computing thermodynamic properties. The ModelSEED biochemistry is continually tested using flux balance analysis to ensure the biochemical network is modeling-ready and capable of simulating diverse phenotypes. Ontologies can be designed to aid in comparing and reconciling metabolic reconstructions that differ in how they represent various metabolic pathways. ModelSEED now includes 33,978 compounds and 36,645 reactions, available as a set of extensible files on GitHub, and available to search at https://modelseed.org/biochem and KBase.

The rapid acceleration of microbial genome sequencing increases opportunities to understand bacterial gene function. Unfortunately, only a small proportion of genes have been studied. Recently, TnSeq has been proposed as a cost-effective, highly reliable approach to predict gene functions as a response to changes in a cell's fitness before-after genomic changes. However, major questions remain about how to best determine whether an observed quantitative change in fitness represents a meaningful change. To address the limitation, we develop a Gaussian mixture model framework for classifying gene function from TnSeq experiments. In order to implement the mixture model, we present the Expectation-Maximization algorithm and a hierarchical Bayesian model sampled using Stan's Hamiltonian Monte-Carlo sampler. We compare these implementations against the frequentist method used in current TnSeq literature. From simulations and real data produced by E. coli TnSeq experiments, we show that the Bayesian implementation of the Gaussian mixture framework provides the most consistent classification results.

Author(s): Arkin, Adam P; Cottingham, Robert W; Henry, Christopher S; Harris, Nomi L; Stevens, Rick L; Maslov, Sergei; Dehal, Paramvir; Ware, Doreen; Perez, Fernando; Canon, Shane; Sneddon, Michael W; Henderson, Matthew L; Riehl, William J; Murphy-Olson, Dan; Chan, Stephen Y; Kamimura, Roy T; Kumari, Sunita; Drake, Meghan M; Brettin, Thomas S; Glass, Elizabeth M; Chivian, Dylan; Gunter, Dan; Weston, David J; Allen, Benjamin H; Baumohl, Jason; Best, Aaron A; Bowen, Ben; Brenner, Steven E; Bun, Christopher C; Chandonia, John-Marc; Chia, Jer-Ming; Colasanti, Ric; Conrad, Neal; Davis, James J; Davison, Brian H; DeJongh, Matthew; Devoid, Scott; Dietrich, Emily; Dubchak, Inna; Edirisinghe, Janaka N; Fang, Gang; Faria, Jose P; Frybarger, Paul M; Gerlach, Wolfgang; Gerstein, Mark; Greiner, Annette; Gurtowski, James; Haun, Holly L; He, Fei; Jain, Rashmi; Joachimiak, Marcin P; Keegan, Kevin P; Kondo, Shinnosuke; Kumar, Vivek; Land, Miriam L; Meyer, Folker; Mills, Marissa; Novichkov, Pavel S; Oh, Taeyun; Olsen, Gary J; Olson, Robert; Parrello, Bruce; Pasternak, Shiran; Pearson, Erik; Poon, Sarah S; Price, Gavin A; Ramakrishnan, Srividya; Ranjan, Priya; Ronald, Pamela C; Schatz, Michael C; Seaver, Samuel MD; Shukla, Maulik; Sutormin, Roman A; Syed, Mustafa H; Thomason, James; Tintle, Nathan L; Wang, Daifeng; Xia, Fangfang; Yoo, Hyunseung; Yoo, Shinjae; Yu, Dantong

An important question in many biological applications, is to estimate or classify gene activity states (active or inactive) based on genome-wide transcriptomics data. Recently, we proposed a Bayesian method, titled MultiMM , which showed superior results compared to existing methods. In short, MultiMM performed better than existing methods on both simulated and real gene expression data, confirming well-known biological results and yielding better agreement with fluxomics data. Despite these promising results, MultiMM has numerous limitations. First, MultiMM leverages co-regulatory models to improve activity state estimates, but information about co-regulation is incorporated in a manner that assumes that networks are known with certainty. Second, MultiMM assumes that genes that change states in the dataset can be distinguished with certainty from those that remain in one state. Third, the model can be sensitive to extreme measures (outliers) of gene expression. In this manuscript, we propose a modified Bayesian approach, which addresses these three limitations by improving outlier handling and by explicitly modeling network and other uncertainty yielding improved gene activity state estimates when compared to MultiMM .

Understanding gene function and regulation is essential for the interpretation, prediction, and ultimate design of cell responses to changes in the environment. An important step toward meeting the challenge of understanding gene function and regulation is the identification of sets of genes that are always co-expressed. These gene sets, Atomic Regulons (ARs), represent fundamental units of function within a cell and could be used to associate genes of unknown function with cellular processes and to enable rational genetic engineering of cellular systems. Here, we describe an approach for inferring ARs that leverages large-scale expression data sets, gene context, and functional relationships among genes. We computed ARs for Escherichia coli based on 907 gene expression experiments and compared our results with gene clusters produced by two prevalent data-driven methods: Hierarchical clustering and k-means clustering. We compared ARs and purely data-driven gene clusters to the curated set of regulatory interactions for E. coli found in RegulonDB, showing that ARs are more consistent with gold standard regulons than are data-driven gene clusters. We further examined the consistency of ARs and data-driven gene clusters in the context of gene interactions predicted by Context Likelihood of Relatedness (CLR) analysis, finding that the ARs show better agreement with CLR predicted interactions. We determined the impact of increasing amounts of expression data on AR construction and find that while more data improve ARs, it is not necessary to use the full set of gene expression experiments available for E. coli to produce high quality ARs. In order to explore the conservation of co-regulated gene sets across different organisms, we computed ARs for Shewanella oneidensis, Pseudomonas aeruginosa, Thermus thermophilus, and Staphylococcus aureus, each of which represents increasing degrees of phylogenetic distance from E. coli. Comparison of the organism-specific ARs showed that the consistency of AR gene membership correlates with phylogenetic distance, but there is clear variability in the regulatory networks of closely related organisms. As large scale expression data sets become increasingly common for model and non-model organisms, comparative analyses of atomic regulons will provide valuable insights into fundamental regulatory modules used across the bacterial domain.

Numerous methods for classifying gene activity states based on gene expression data have been proposed for use in downstream applications, such as incorporating transcriptomics data into metabolic models in order to improve resulting flux predictions. These methods often attempt to classify gene activity for each gene in each experimental condition as belonging to one of two states: active (the gene product is part of an active cellular mechanism) or inactive (the cellular mechanism is not active). These existing methods of classifying gene activity states suffer from multiple limitations, including enforcing unrealistic constraints on the overall proportions of active and inactive genes, failing to leverage a priori knowledge of gene co-regulation, failing to account for differences between genes, and failing to provide statistically meaningful confidence estimates. We propose a flexible Bayesian approach to classifying gene activity states based on a Gaussian mixture model. The model integrates genome-wide transcriptomics data from multiple conditions and information about gene co-regulation to provide activity state confidence estimates for each gene in each condition. We compare the performance of our novel method to existing methods on both simulated data and real data from 907 E. coli gene expression arrays, as well as a comparison with experimentally measured flux values in 29 conditions, demonstrating that our method provides more consistent and accurate results than existing methods across a variety of metrics.

With continued rapid growth in the number and quality of fully sequenced and accurately annotated bacterial genomes, we have unprecedented opportunities to understand metabolic diversity. We selected 101 diverse and representative completely sequenced bacteria and implemented a manual curation effort to identify 846 unique metabolic variants present in these bacteria. The presence or absence of these variants act as a metabolic signature for each of the bacteria, which can then be used to understand similarities and differences between and across bacterial groups. We propose a novel and robust method of summarizing metabolic diversity using metabolic signatures and use this method to generate a metabolic tree, clustering metabolically similar organisms. Resulting analysis of the metabolic tree confirms strong associations with well-established biological results along with direct insight into particular metabolic variants which are most predictive of metabolic diversity. The positive results of this manual curation effort and novel method development suggest that future work is needed to further expand the set of bacteria to which this approach is applied and use the resulting tree to test broad questions about metabolic diversity and complexity across the bacterial tree of life.

BACKGROUND:Rapid growth in the availability of genome-wide transcript abundance levels through gene expression microarrays and RNAseq promises to provide deep biological insights into the complex, genome-wide transcriptional behavior of single-celled organisms. However, this promise has not yet been fully realized.RESULTS:We find that computation of pairwise gene associations (correlation; mutual information) across a set of 2782 total genome-wide expression samples from six diverse bacteria produces unexpectedly large variation in estimates of pairwise gene association-regardless of the metric used, the organism under study, or the number and source of the samples. We pinpoint the cause to sampling bias. In particular, in repositories of expression data (e.g., Gene Expression Omnibus, GEO), many individual genes show small differences in absolute gene expression levels across the set of samples. We demonstrate that these small differences are due mainly to "noise" instead of "signal" attributable to environmental or genetic perturbations. We show that downstream analysis using gene expression levels of genes with small differences yields biased estimates of pairwise association.CONCLUSIONS:We propose flagging genes with small differences in absolute, RMA-normalized, expression levels (e.g., standard deviation less than 0.5), as potentially yielding biased pairwise association metrics. This strategy has the potential to substantially improve the confidence in genome-wide conclusions about transcriptional behavior in bacterial organisms. Further work is needed to further refine strategies to identify genes with small difference in expression levels prior to computing gene-gene association metrics.

Over the past decade, genome-scale metabolic models have proven to be a crucial resource for predicting organism phenotypes from genotypes. These models provide a means of rapidly translating detailed knowledge of thousands of enzymatic processes into quantitative predictions of whole-cell behavior. Until recently, the pace of new metabolic model development was eclipsed by the pace at which new genomes were being sequenced. To address this problem, the RAST and the Model SEED framework were developed as a means of automatically producing annotations and draft genome-scale metabolic models. In this chapter, we describe the automated model reconstruction process in detail, starting from a new genome sequence and finishing on a functioning genome-scale metabolic model. We break down the model reconstruction process into eight steps: submitting a genome sequence to RAST, annotating the genome, curating the annotation, submitting the annotation to Model SEED, reconstructing the core model, generating the draft biomass reaction, auto-completing the model, and curating the model. Each of these eight steps is documented in detail.

We have developed a Cytoscape plugin for exploring genome‐scale metabolic models in the Model SEED database (www.theseed.org/models). We have used this plugin to compare our metabolic model for Mycoplasma pneumoniae M129 (Mp) with the one published by Yus et al. in Science in 2009. Mp is missing an annotated gene for the transaldolase (TAL) activity, which converts erythrose‐4‐phosphate into sedoheptulose‐7‐phosphate. However, it is known that Mp produces phospho‐ribose‐diphosphate, indicating that the pentose phosphate pathway is functional. The Yus model asserts that the TAL activity is present; however, our software shows that two reactions present in our model associated with fructose‐bisphosphate aldolase and phosphofructokinase bridge this gap using sedoheptulose‐1,7‐bisphosphate as an intermediary compound. Our software also shows differences in asserted transport reactions, glycerophospholipid metabolism, and maintenance of oxidation‐reduction balance. Our software enables comparing the results of flux variability analysis for the two models on different growth media, and the determination of which reactions are essential on the minimal media discovered by Yus et al.NSF 0745100, 0850546

BACKGROUND:Statistical analyses of whole genome expression data require functional information about genes in order to yield meaningful biological conclusions. The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) are common sources of functionally grouped gene sets. For bacteria, the SEED and MicrobesOnline provide alternative, complementary sources of gene sets. To date, no comprehensive evaluation of the data obtained from these resources has been performed.RESULTS:We define a series of gene set consistency metrics directly related to the most common classes of statistical analyses for gene expression data, and then perform a comprehensive analysis of 3581 Affymetrix® gene expression arrays across 17 diverse bacteria. We find that gene sets obtained from GO and KEGG demonstrate lower consistency than those obtained from the SEED and MicrobesOnline, regardless of gene set size.CONCLUSIONS:Despite the widespread use of GO and KEGG gene sets in bacterial gene expression data analysis, the SEED and MicrobesOnline provide more consistent sets for a wide variety of statistical analyses. Increased use of the SEED and MicrobesOnline gene sets in the analysis of bacterial gene expression data may improve statistical power and utility of expression data.

CytoSEED is a Cytoscape plugin for viewing, manipulating and analyzing metabolic models created using the Model SEED. The CytoSEED plugin enables users of the Model SEED to create informative visualizations of the reaction networks generated for their organisms of interest. These visualizations are useful for understanding organism-specific biochemistry and for highlighting the results of flux variability analysis experiments.

ABSTRACT Transcriptional regulatory networks are fine-tuned systems that help microorganisms respond to changes in the environment and cell physiological state. We applied the comparative genomics approach implemented in the RegPredict Web server combined with SEED subsystem analysis and available information on known regulatory interactions for regulatory network reconstruction for the human pathogen Staphylococcus aureus and six related species from the family Staphylococcaceae . The resulting reference set of 46 transcription factor regulons contains more than 1,900 binding sites and 2,800 target genes involved in the central metabolism of carbohydrates, amino acids, and fatty acids; respiration; the stress response; metal homeostasis; drug and metal resistance; and virulence. The inferred regulatory network in S. aureus includes ∼320 regulatory interactions between 46 transcription factors and ∼550 candidate target genes comprising 20% of its genome. We predicted ∼170 novel interactions and 24 novel regulons for the control of the central metabolic pathways in S. aureus . The reconstructed regulons are largely variable in the Staphylococcaceae : only 20% of S. aureus regulatory interactions are conserved across all studied genomes. We used a large-scale gene expression data set for S. aureus to assess relationships between the inferred regulons and gene expression patterns. The predicted reference set of regulons is captured within the Staphylococcus collection in the RegPrecise database ( http://regprecise.lbl.gov ).

Background: The development of next generation sequencing technology is rapidly changing the face of the genome annotation and analysis field. One of the primary uses for genome sequence data is to improve our understanding and prediction of phenotypes for microbes and microbial communities, but the technologies for predicting phenotypes must keep pace with the new sequences emerging.Scope of review: This review presents an integrated view of the methods and technologies used in the inference of phenotypes for microbes and microbial communities based on genomic and metagenomic data. Given the breadth of this topic, we place special focus on the resources available within the SEED Project. We discuss the two steps involved in connecting genotype to phenotype: sequence annotation, and phenotype inference, and we highlight the challenges in each of these steps when dealing with both single genome and metagenome data.Major conclusions: This integrated view of the genotype-to-phenotype problem highlights the importance of a controlled ontology in the annotation of genomic data, as this benefits subsequent phenotype inference and metagenome annotation. We also note the importance of expanding the set of reference genomes to improve the annotation of all sequence data, and we highlight metagenome assembly as a potential new source for complete genomes. Finally, we find that phenotype inference, particularly from metabolic models, generates predictions that can be validated and reconciled to improve annotations.General significance: This review presents the first look at the challenges and opportunities associated with the inference of phenotype from genotype during the next generation sequencing revolution. This article is part of a Special Issue entitled: Systems Biology of Microorganisms. Published by Elsevier BM.

The breadth-first approach to teaching introductory computer science is one way of dispelling the common misperception that programming is the sole task of the computer scientist. The breadth-first approach is particularly useful in courses for non-majors. Hands-on activities that make up laboratory assignments for these courses tend to focus on learning to program or simulations of program execution. These activities unfortunately fail to build on the foundations laid by a breadth-first approach, and serve to perpetuate the computer science = programming misperception. We have developed a set of laboratory activities which are based on what we call authentic computing tasks: everyday tasks that students want to know how to accomplish. Example tasks include image editing, operating system installation and configuration, and building home computer networks. Explicit connections are made between these authentic computing tasks and the computer science concepts being covered in the lecture portion of the course. The course has experienced dramatic increases in enrollment, and we have evidence that students see the connections, rather than coming to believe that performing computing tasks well is the essence of computer science.

Reconstructing a metabolic model from the genome sequence of an organism is a useful but arduous approach for predicting phenotypes. Henry et al . describe a resource that automates most of this process and apply it to create >100 new metabolic models of microbes.

We have recently developed a high‐throughput computational pipeline for generating genome‐scale metabolic models. To test this pipeline, we have generated models for three mollicutes: Mycoplasma genitalium G‐37 (Mg), Ureaplasma parvum serovar 3 (Up), and Mycoplasma hominis (Mh). All three genomes lack an annotated transaldolase (TAL) gene, resulting in incomplete pentose phosphate pathways (PPP). However, all three organisms require production of phospho‐ribose‐diphosphate (PRPP), a key product of the PPP. To meet this demand, all three models predict novel routes through the PPP. The route predicted in the Mg and Up models relies on secondary activities of the phosphofructokinase (PFK) and fructose‐bisphosphate aldolase (FBA) enzymes to cope with the accumulation of sedoheptulose 7‐phosphate. This PPP bypass was recently discovered in an E. coli mutant lacking the TAL gene. Because Mh lacks a gene for PFK, the Mh model predicts a different route, reversing the activity of FBA and postulating the presence of Sedoheptulose‐1,7‐bisphosphatase. There is no genomic evidence for the presence of known forms of this enzyme in Mh; in this case, the model reveals the necessity of clarifying the pathway of PRPP production in Mh. Our study highlights the potential for comparative analysis of metabolic models to predict novel biochemistry and suggest testable hypotheses in prokaryotic metabolism.NSF MCB 0745100