Dataset Information

Predicting interpretability of metabolome models based on behavior, putative identity, and biological relevance of explanatory signals.

ABSTRACT: Powerful algorithms are required to deal with the dimensionality of metabolomics data. Although many achieve high classification accuracy, the models they generate have limited value unless it can be demonstrated that they are reproducible and statistically relevant to the biological problem under investigation. Random forest (RF) generates models, without any requirement for dimensionality reduction or feature selection, in which individual variables are ranked for significance and displayed in an explicit manner. In metabolome fingerprinting by mass spectrometry, each metabolite can be represented by signals at several m/z. Exploiting a prior understanding of expected biochemical differences between sample classes, we aimed to develop meaningful metrics relevant to the significance both of the overall RF model and individual, potentially explanatory, signals. Pair-wise comparison of related plant genotypes with strong phenotypic differences demonstrated that robust models are not only reproducible but also logically structured, highlighting correlated m/z derived from just a small number of explanatory metabolites reflecting the biological differences between sample classes. RF models were also generated by using groupings of samples known to be increasingly phenotypically similar. Although classification accuracy was often reasonable, we demonstrated reproducibly in both Arabidopsis and potato a performance threshold based on margin statistics beyond which such models showed little structure indicative of either generalizability or further biological interpretability. In a multiclass problem using 25 Arabidopsis genotypes, despite the complicating effects of ecotype background and secondary metabolome perturbations common to several mutations, the ranking of metabolome signals by RF provided scope for deeper interpretability.

SUBMITTER: Enot DP

PROVIDER: S-EPMC1595442 | biostudies-literature | 2006 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Predicting interpretability of metabolome models based on behavior, putative identity, and biological relevance of explanatory signals.

Enot David P DP Beckmann Manfred M Overy David D Draper John J

Proceedings of the National Academy of Sciences of the United States of America 20060921 40

Powerful algorithms are required to deal with the dimensionality of metabolomics data. Although many achieve high classification accuracy, the models they generate have limited value unless it can be demonstrated that they are reproducible and statistically relevant to the biological problem under investigation. Random forest (RF) generates models, without any requirement for dimensionality reduction or feature selection, in which individual variables are ranked for significance and displayed in ...[more]

PMID: 16990432

Similar Datasets

Project description:Satellite tobacco necrosis virus 1 (STNV-1) is a model system for in vitro RNA encapsidation studies (N. Patel, E. C. Dykeman, R. H. A. Coutts, G. P. Lomonossoff, et al., Proc Natl Acad Sci U S A 112:2227-2232, 2015, https://doi.org/10.1073/pnas.1420812112; N. Patel, E. Wroblewski, G. Leonov, S. E. V. Phillips, et al., Proc Natl Acad Sci U S A 114:12255-12260, 2017, https://doi.org/10.1073/pnas.1706951114), leading to the identification of degenerate packaging signals (PSs) proposed to be involved in the recognition of its genome by the capsid protein (CP). The aim of the present work was to investigate whether these putative PSs can confer selective packaging of STNV-1 RNA in vivo and to assess the prospects of using decoy RNAs in antiviral therapy. We have developed an in planta packaging assay based on the transient expression of STNV-1 CP and have assessed the ability of the resulting virus-like particles (VLPs) to encapsidate mutant STNV-1 RNAs expected to have different encapsidation potential based on in vitro studies. The results revealed that >90% of the encapsidated RNAs are host derived, although there is some selectivity of packaging for STNV-1 RNA and certain host RNAs. Comparison of the packaging efficiencies of mutant STNV-1 RNAs showed that they are encapsidated mainly according to their abundance within the cells, rather than the presence or absence of the putative PSs previously identified from in vitro studies. In contrast, subsequent infection experiments demonstrated that host RNAs represent only <1% of virion content. Although selective encapsidation of certain host RNAs was noted, no direct correlation could be made between this preference and the presence of potential PSs in the host RNA sequences. Overall, the data illustrate that the differences in RNA packaging efficiency identified through in vitro studies are insufficient to explain the specific packaging of STNV-1 RNA.IMPORTANCE Viruses preferentially encapsidate their own genomic RNA, sometimes as a result of the presence of clearly defined packaging signals (PSs) in their genome sequence. Recently, a novel form of short degenerate PSs has been proposed (N. Patel, E. C. Dykeman, R. H. A. Coutts, G. P. Lomonossoff, et al., Proc Natl Acad Sci U S A 112:2227-2232, 2015, https://doi.org/10.1073/pnas.1420812112; N. Patel, E. Wroblewski, G. Leonov, S. E. V. Phillips, et al., Proc Natl Acad Sci U S A 114:12255-12260, 2017, https://doi.org/10.1073/pnas.1706951114) using satellite tobacco necrosis virus 1 (STNV-1) as a model system for in vitro studies. It has been suggested that competing with these putative PSs may constitute a novel therapeutic approach against pathogenic single-stranded RNA viruses. Our work demonstrates that the previously identified PSs have no discernible significance for the selective packaging of STNV-1 in vivo in the presence and absence of competition or replication: viral sequences are encapsidated mostly on the basis of their abundance within the cell, while encapsidation of host RNAs also occurs. Nevertheless, the putative PSs identified in STNV-1 RNA may still have applications in bionanotechnology, such as the in vitro selective packaging of RNA molecules.

Project description:Univariate analyses of metabolomics data currently follow a frequentist approach, using p-values to reject a null hypothesis. We here propose the use of Bayesian statistics to quantify evidence supporting different hypotheses and discriminate between the null hypothesis versus the lack of statistical power. We used metabolomics data from three independent human cohorts that studied the plasma signatures of subjects with myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS). The data are publicly available, covering 84-197 subjects in each study with 562-888 identified metabolites of which 777 were common between the two studies and 93 were compounds reported in all three studies. We show how Bayesian statistics incorporates results from one study as "prior information" into the next study, thereby improving the overall assessment of the likelihood of finding specific differences between plasma metabolite levels. Using classic statistics and Benjamini-Hochberg FDR-corrections, Study 1 detected 18 metabolic differences and Study 2 detected no differences. Using Bayesian statistics on the same data, we found a high likelihood that 97 compounds were altered in concentration in Study 2, after using the results of Study 1 as the prior distributions. These findings included lower levels of peroxisome-produced ether-lipids, higher levels of long-chain unsaturated triacylglycerides, and the presence of exposome compounds that are explained by the difference in diet and medication between healthy subjects and ME/CFS patients. Although Study 3 reported only 92 compounds in common with the other two studies, these major differences were confirmed. We also found that prostaglandin F2alpha, a lipid mediator of physiological relevance, was reduced in ME/CFS patients across all three studies. The use of Bayesian statistics led to biological conclusions from metabolomic data that were not found through frequentist approaches. We propose that Bayesian statistics is highly useful for studies with similar research designs if similar metabolomic assays are used.

Dataset Information

Predicting interpretability of metabolome models based on behavior, putative identity, and biological relevance of explanatory signals.

Publications

Predicting interpretability of metabolome models based on behavior, putative identity, and biological relevance of explanatory signals.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets