Choosing where to look next in a mutation sequence space: Active Learning of informative p53 cancer rescue mutants

MOTIVATION: Many biomedical projects would benefit from reducing the time and expense of in vitro experimentation by using computer models for in silico predictions. These models may help determine which expensive biological data are most useful to acquire next. Active Learning techniques for choosing the most informative data enable biologists and computer scientists to optimize experimental data choices for rapid discovery of biological function. To explore design choices that affect this desirable behavior, five novel and five existing Active Learning techniques, together with three control methods, were tested on 57 previously unknown p53 cancer rescue mutants for their ability to build classifiers that predict protein function. The best of these techniques, Maximum Curiosity, improved the baseline accuracy of 56-77%. This article shows that Active Learning is a useful tool for biomedical research, and provides a case study of interest to others facing similar discovery challenges.     

Computational and experimental analysis of redundancy in the central metabolism of Geobacter sulfurreducens

Previous model-based analysis of the metabolic network of Geobacter sulfurreducens suggested the existence of several redundant pathways. Here, we identified eight sets of redundant pathways that included redundancy for the assimilation of acetate, and for the conversion of pyruvate into acetyl-CoA. These equivalent pathways and two other sub-optimal pathways were studied using 5 single-gene deletion mutants in those pathways for the evaluation of the predictive capacity of the model. The growth phenotypes of these mutants were studied under 12 different conditions of electron donor and acceptor availability. The comparison of the model predictions with the resulting experimental phenotypes indicated that pyruvate ferredoxin oxidoreductase is the only activity able to convert pyruvate into acetyl-CoA. However, the results and the modeling showed that the two acetate activation pathways present are not only active, but needed due to the additional role of the acetyl-CoA transferase in the TCA cycle, probably reflecting the adaptation of these bacteria to acetate utilization. In other cases, the data reconciliation suggested additional capacity constraints that were confirmed with biochemical assays. The results demonstrate the need to experimentally verify the activity of key enzymes when developing in silico models of microbial physiology based on sequence-based reconstruction of metabolic networks.     

Machine Learning Techniques for the Automated Classification of Adhesin-Like Proteins in the Human Protozoan Parasite Trypanosoma cruzi

This paper reports on the evaluation of different machine learning techniques for the automated classification of coding gene sequences obtained from several organisms in terms of their functional role as adhesins. Diverse, biologically-meaningful, sequence-based features were extracted from the sequences and used as inputs to the in silico prediction models. Another contribution of this work is the generation of potentially novel and testable predictions about the surface protein DGF-1 family in Trypanosoma cruzi. Finally, these techniques are potentially useful for the automated annotation of known adhesin-like proteins from the trans-sialidase surface protein family in T. cruzi, the etiological agent of Chagas disease.     

Comparative metabolic capabilities for Micrococcus luteus NCTC 2665, the "Fleming" strain, and actinobacteria

Putative gene predictions of the Gram positive actinobacteria Micrococcus luteus (NCTC 2665, "Fleming strain") was used to construct a genome scale reconstruction of the metabolic network for this organism. The metabolic network comprises 586 reactions and 551 metabolites, and accounts for 21% of the genes in the genome. The reconstruction was based on the annotated genome and available biochemical information. M. luteus has one of the smallest genomes of actinobacteria with a circular chromosome of 2,501,097 base pairs and a GC content of 73%. The metabolic pathways required for biomass production in silico were determined based on earlier models of actinobacteria. The in silico network is used for metabolic comparison of M. luteus with other actinomycetes, and hence provides useful information for possible future biotechnological exploitation of this organism, e.g., for production of biofuels.     

Improving the species cross-reactivity of an antibody using computational design

The high degree of specificity displayed by antibodies often results in varying potencies against antigen orthologs, which can affect the efficacy of these molecules in different animal models of disease. We have used a computational design strategy to improve the species cross-reactivity of an antibody-based inhibitor of the cancer-associated serine protease MT-SP1. In silico predictions were tested in vitro, and the most effective mutation, T98R, was shown to improve antibody affinity for the mouse ortholog of the enzyme 14-fold, resulting in an inhibitor with a K_I of 340 pM. This improved affinity will be valuable when exploring the role of MT-SP1 in mouse models of cancer, and the strategy outlined here could be useful in fine-tuning antibody specificity.     

Effects of Genetic and Environmental Factors on Trait Network Predictions From Quantitative Trait Locus Data

The use of high-throughput genomic techniques to map gene expression quantitative trait loci has spurred the development of path analysis approaches for predicting functional networks linking genes and natural trait variation. The goal of this study was to test whether potentially confounding factors, including effects of common environment and genes not included in path models, affect predictions of cause–effect relationships among traits generated by QTL path analyses. Structural equation modeling (SEM) was used to test simple QTL-trait networks under different regulatory scenarios involving direct and indirect effects. SEM identified the correct models under simple scenarios, but when common-environment effects were simulated in conjunction with direct QTL effects on traits, they were poorly distinguished from indirect effects, leading to false support for indirect models. Application of SEM to loblolly pine QTL data provided support for biologically plausible a priori hypotheses of QTL mechanisms affecting height and diameter growth. However, some biologically implausible models were also well supported. The results emphasize the need to include any available functional information, including predictions for genetic and environmental correlations, to develop plausible models if biologically useful trait network predictions are to be made.     

Network-based assessment of the selectivity of metabolic drug targets in Plasmodium falciparum with respect to human liver metabolism

ABSTRACT: BACKGROUND: The search for new drug targets for antibiotics against Plasmodium falciparum, a major cause of human deaths, is a pressing scientific issue, as multiple resistance strains spread rapidly. Metabolic network-based analyses may help to identify those parasite's essential enzymes whose homologous counterparts in the human host cells are either absent, non-essential or relatively less essential. RESULTS: Using the well-curated metabolic networks PlasmoNet of the parasite Plasmodium falciparum and HepatoNet1 of the human hepatocyte, the selectivity of 48 experimental antimalarial drug targets was analyzed. Applying in silico gene deletions, 24 of these drug targets were found to be perfectly selective, in that they were essential for the parasite but non-essential for the human cell. The selectivity of a subset of enzymes, that were essential in both models, was evaluated with the reduced fitness concept. It was, then, possible to quantify the reduction in functional fitness of the two networks under the progressive inhibition of the same enzymatic activity. Overall, this in silico analysis provided a selectivity ranking that was in line with numerous in vivo and in vitro observations. CONCLUSIONS: Genome-scale models can be useful to depict and quantify the effects of enzymatic inhibitions on the impaired production of biomass components. From the perspective of a host-pathogen metabolic interaction, an estimation of the drug targets-induced consequences can be beneficial for the development of a selective anti-parasitic drug.     

神经网络与人事决策研究的新范式

随着多特征决策研究的深入,传统方法已经不能回答更加细致的问题。细察精确预测的理论、建立模型与数据的形式关系成为更有希望的研究方向。神经网络模型设计用来模拟许多并行的认知和神经行为,具有样例学习和迁移适应能力．在解释和预测方面具有传统方法所不具备的潜力。神经网络能够同时表征线性补偿和非补偿规则,其应用已经渗透到许多学科领域。网络范式对于人事研究和应用也有价值,有研究表明神经网络可以用于人力资源管理的一般领域,成为人事决策研究的新范式。     

Mutualistic networks: moving closer to a predictive theory

Plant–animal mutualistic networks sustain terrestrial biodiversity and human food security. Global environmental changes threaten these networks, underscoring the urgency for developing a predictive theory on how networks respond to perturbations. Here, I synthesise theoretical advances towards predicting network structure, dynamics, interaction strengths and responses to perturbations. I find that mathematical models incorporating biological mechanisms of mutualistic interactions provide better predictions of network dynamics. Those mechanisms include trait matching, adaptive foraging, and the dynamic consumption and production of both resources and services provided by mutualisms. Models incorporating species traits better predict the potential structure of networks (fundamental niche), while theory based on the dynamics of species abundances, rewards, foraging preferences and reproductive services can predict the extremely dynamic realised structures of networks, and may successfully predict network responses to perturbations. From a theoretician's standpoint, model development must more realistically represent empirical data on interaction strengths, population dynamics and how these vary with perturbations from global change. From an empiricist's standpoint, theory needs to make specific predictions that can be tested by observation or experiments. Developing models using short‐term empirical data allows models to make longer term predictions of community dynamics. As more longer term data become available, rigorous tests of model predictions will improve.     

Potential drug targets on insomnia and intervention effects of Jujuboside A through metabolic pathway analysis as revealed by UPLC/ESI-SYNAPT-HDMS coupled with pattern recognition approach

Potential metabolites from the metabolic pathways could be therapeutic targets and useful for the discovery of broad spectrum drugs. UPLC/ESI-SYNAPT-HDMS coupled with pattern recognition methods including PCA, PLS-DA, OPLS-DA and Heatmap were integrated to examine the global metabolic signature of insomnia and intervention effects of Jujuboside A (JuA). Six unique pathways of the insomnia were identified using Ingenuity Pathway Analysis (IPA) software. The VIP-value threshold cutoff of the metabolites was set to 10, above this threshold, were filtered out as potential target biomarkers. Sixteen distinct metabolites were identified from these pathways, and 6 of them can be considered for rational drug design. It was further experimental validation that the changes in metabolic profiling were restored to their baseline values after JuA treatment according to the multivariate data analysis. Potential metabolite network of the insomnia was preliminarily predicted JuA-target interaction networks, and could be further explored for in silico docking studies with suitable drugs. Thus, our method is an efficient procedure for drug target identification through metabolic analysis. It can guide testable predictions, provide insights into drug action mechanisms and enable us to increase research productivity toward metabolomic drug discovery.     

On the edge: modeling protrusion

Actin-based protrusion is the first step in cell crawling. In the last two decades, the studies of actin networks in the lamellipodium and Listeria's comet tail advanced so far that the last goal of the reductionist agenda - reconstitution of protrusion from purified components in vitro and in silico - became viable. Earlier models dealt with growth of and force generation by a single actin filament. Modern models of tethered ratchet, autocatalytic branching, end-tracking motor action and elastic- and nano- propulsion have recently helped to elucidate dynamics and forces in complex actin networks. By considering these models, their limitations and their relationships to recent biophysical data, progress is being made toward a unified model of protrusion.     

Novel insights through the integration of structural and functional genomics data with protein networks

In recent years, major advances in genomics, proteomics, macromolecular structure determination, and the computational resources capable of processing and disseminating the large volumes of data generated by each have played major roles in advancing a more systems-oriented appreciation of biological organization. One product of systems biology has been the delineation of graph models for describing genome-wide protein-protein interaction networks. The network organization and topology which emerges in such models may be used to address fundamental questions in an array of cellular processes, as well as biological features intrinsic to the constituent proteins (or "nodes") themselves. However, graph models alone constitute an abstraction which neglects the underlying biological and physical reality that the network's nodes and edges are highly heterogeneous entities. Here, we explore some of the advantages of introducing a protein structural dimension to such models, as the marriage of conventional network representations with macromolecular structural data helps to place static node and edge constructs in a biologically more meaningful context. We emphasize that 3D protein structures constitute a valuable conceptual and predictive framework by discussing examples of the insights provided, such as enabling in silico predictions of protein-protein interactions, providing rational and compelling classification schemes for network elements, as well as revealing interesting intrinsic differences between distinct node types, such as disorder and evolutionary features, which may then be rationalized in light of their respective functions within networks.     

The statistical mechanics of complex signaling networks: nerve growth factor signaling

The inherent complexity of cellular signaling networks and their importance to a wide range of cellular functions necessitates the development of modeling methods that can be applied toward making predictions and highlighting the appropriate experiments to test our understanding of how these systems are designed and function. We use methods of statistical mechanics to extract useful predictions for complex cellular signaling networks. A key difficulty with signaling models is that, while significant effort is being made to experimentally measure the rate constants for individual steps in these networks, many of the parameters required to describe their behavior remain unknown or at best represent estimates. To establish the usefulness of our approach, we have applied our methods toward modeling the nerve growth factor (NGF)-induced differentiation of neuronal cells. In particular, we study the actions of NGF and mitogenic epidermal growth factor (EGF) in rat pheochromocytoma (PC12) cells. Through a network of intermediate signaling proteins, each of these growth factors stimulates extracellular regulated kinase (Erk) phosphorylation with distinct dynamical profiles. Using our modeling approach, we are able to predict the influence of specific signaling modules in determining the integrated cellular response to the two growth factors. Our methods also raise some interesting insights into the design and possible evolution of cellular systems, highlighting an inherent property of these systems that we call 'sloppiness.'     

Identification of genome-scale metabolic network models using experimentally measured flux profiles

Genome-scale metabolic network models can be reconstructed for well-characterized organisms using genomic annotation and literature information. However, there are many instances in which model predictions of metabolic fluxes are not entirely consistent with experimental data, indicating that the reactions in the model do not match the active reactions in the in vivo system. We introduce a method for determining the active reactions in a genome-scale metabolic network based on a limited number of experimentally measured fluxes. This method, called optimal metabolic network identification (OMNI), allows efficient identification of the set of reactions that results in the best agreement between in silico predicted and experimentally measured flux distributions. We applied the method to intracellular flux data for evolved Escherichia coli mutant strains with lower than predicted growth rates in order to identify reactions that act as flux bottlenecks in these strains. The expression of the genes corresponding to these bottleneck reactions was often found to be downregulated in the evolved strains relative to the wild-type strain. We also demonstrate the ability of the OMNI method to diagnose problems in E. coli strains engineered for metabolite overproduction that have not reached their predicted production potential. The OMNI method applied to flux data for evolved strains can be used to provide insights into mechanisms that limit the ability of microbial strains to evolve towards their predicted optimal growth phenotypes. When applied to industrial production strains, the OMNI method can also be used to suggest metabolic engineering strategies to improve byproduct secretion. In addition to these applications, the method should prove to be useful in general for reconstructing metabolic networks of ill-characterized microbial organisms based on limited amounts of experimental data.     

Biochemical and statistical network models for systems biology

The normal and abnormal behavior of a living cell is governed by complex networks of interacting biomolecules. Models of these networks allow us to make predictions about cellular behavior under a variety of environmental cues. In this review, we focus on two broad classes of such models: biochemical network models and statistical inference models. In particular, we discuss a number of modeling approaches in the context of the assumptions that they entail, the types of data required for their inference, and the range of their applicability.