Missing value imputation strategies for metabolomics data

Armitage, Emily G.; Godzień, Joanna; Alonso-Herranz, Vanesa; López‐Gonzálvez, Ángeles; Barbas, Coral

doi:10.1002/elps.201500352

Cited by 132 publications

(112 citation statements)

References 23 publications

Supporting

Mentioning

109

Contrasting

Unclassified

Order By: Relevance

“…There are different methods for dealing with missing values, and each method/approach impacts on downstream statistical analyses [31,50,51]. In the present study, the SIMCA software uses an adjusted nonlinear iterative partial least squares (NIPALS) algorithm (with a correction factor of 3.0) [52] in handling the missing values.…”

Section: Resultsmentioning

confidence: 99%

A Conversation on Data Mining Strategies in LC-MS Untargeted Metabolomics: Pre-Processing and Pre-Treatment Steps

et al. 2016

View full text Add to dashboard Cite

Untargeted metabolomic studies generate information-rich, high-dimensional, and complex datasets that remain challenging to handle and fully exploit. Despite the remarkable progress in the development of tools and algorithms, the “exhaustive” extraction of information from these metabolomic datasets is still a non-trivial undertaking. A conversation on data mining strategies for a maximal information extraction from metabolomic data is needed. Using a liquid chromatography-mass spectrometry (LC-MS)-based untargeted metabolomic dataset, this study explored the influence of collection parameters in the data pre-processing step, scaling and data transformation on the statistical models generated, and feature selection, thereafter. Data obtained in positive mode generated from a LC-MS-based untargeted metabolomic study (sorghum plants responding dynamically to infection by a fungal pathogen) were used. Raw data were pre-processed with MarkerLynxTM software (Waters Corporation, Manchester, UK). Here, two parameters were varied: the intensity threshold (50–100 counts) and the mass tolerance (0.005–0.01 Da). After the pre-processing, the datasets were imported into SIMCA (Umetrics, Umea, Sweden) for more data cleaning and statistical modeling. In addition, different scaling (unit variance, Pareto, etc.) and data transformation (log and power) methods were explored. The results showed that the pre-processing parameters (or algorithms) influence the output dataset with regard to the number of defined features. Furthermore, the study demonstrates that the pre-treatment of data prior to statistical modeling affects the subspace approximation outcome: e.g., the amount of variation in X-data that the model can explain and predict. The pre-processing and pre-treatment steps subsequently influence the number of statistically significant extracted/selected features (variables). Thus, as informed by the results, to maximize the value of untargeted metabolomic data, understanding of the data structures and exploration of different algorithms and methods (at different steps of the data analysis pipeline) might be the best trade-off, currently, and possibly an epistemological imperative.

show abstract

Section: Resultsmentioning

confidence: 99%

A Conversation on Data Mining Strategies in LC-MS Untargeted Metabolomics: Pre-Processing and Pre-Treatment Steps

et al. 2016

View full text Add to dashboard Cite

show abstract

“…The missing data were imputed by k-nearest neighbours algorithm combined with linear regression technique [53]. The proteins with significant fold change greater than 2 or lower than 0.5 were recognised as deregulated.…”

Section: Methodsmentioning

confidence: 99%

Transcriptomic and proteomic analysis of mouse radiation-induced acute myeloid leukaemia (AML)

et al. 2016

View full text Add to dashboard Cite

A combined transcriptome and proteome analysis of mouse radiation-induced AMLs using two primary AMLs, cell lines from these primaries, another cell line and its in vivo passage is reported. Compared to haematopoietic progenitor and stem cells (HPSC), over 5000 transcriptome alterations were identified, 2600 present in all materials. 55 and 3 alterations were detected in the proteomes of the cell lines and primary/in vivo passage material respectively, with one common to all materials. In cell lines, approximately 50% of the transcriptome changes are related to adaptation to cell culture, and in the proteome this proportion was higher. An AML ‘signature’ of 17 genes/proteins commonly deregulated in primary AMLs and cell lines compared to HPSCs was identified and validated using human AML transcriptome data. This also distinguishes primary AMLs from cell lines and includes proteins such as Coronin 1, pontin/RUVBL1 and Myeloperoxidase commonly implicated in human AML. C-Myc was identified as having a key role in radiation leukaemogenesis. These data identify novel candidates relevant to mouse radiation AML pathogenesis, and confirm that pathways of leukaemogenesis in the mouse and human share substantial commonality.

show abstract

“…From a theoretical point of view, this low percentage is rather generous as the instrumentation should be sufficiently reliable to detect a larger percentage of features in each of the repeated injections of QC sample. Further indepth reading-recommendations relevant to both the possible reasons for missing values in metabolomics datasets as well as adequate solutions include articles by Gromski et al (Gromski et al 2014) and Armitage et al (Armitage et al 2015). Further refinement of this approach include analysis of a QC dilution series at the end of the analysis to identify features not responsive to dilution, these should then be excluded (Eliasson et al 2012;Vorkas et al 2015).…”

Section: Evaluation Of Data Qualitymentioning

confidence: 96%

LC–MS based global metabolite profiling: the necessity of high data quality

et al. 2016

View full text Add to dashboard Cite

LC-MS based global metabolite profiling currently lacks detailed guidelines to demonstrate that the obtained data is of high enough analytical quality. Insufficient data quality may result in the failure to generate a hypothesis, or in the worst case, a false or skewed hypothesis. After assessing the literature, it is apparent that an analytically focused summary and critical discussion related to this subject would be beneficial for both beginners and experts engaged in this field. A particular focus will be placed on data quality, which we here define as the degree to which a set of parameters fulfills predetermined criteria, similar to the established guidelines for targeted analysis. However, several of these parameters are difficult to assess since holistic approaches measure thousands of metabolites in parallel and seldom include predefined knowledge of which metabolites will differ between sample groups. In this review, the following parameters will be discussed in detail: reproducibility, selectivity, certainty of metabolite identification and metabolite coverage. The review systematically describes the generic workflow for LC-MS based global metabolite profiling and highlights how each separate part may affect data quality. The last part of the review describes how data quality can be evaluated as well as identifies areas where additional improvement is needed. In this review, we provide our own analytical opinions in regards to evaluation and, to some extent, improvement of data quality.

show abstract

Missing value imputation strategies for metabolomics data

Cited by 132 publications

References 23 publications

A Conversation on Data Mining Strategies in LC-MS Untargeted Metabolomics: Pre-Processing and Pre-Treatment Steps

A Conversation on Data Mining Strategies in LC-MS Untargeted Metabolomics: Pre-Processing and Pre-Treatment Steps

Transcriptomic and proteomic analysis of mouse radiation-induced acute myeloid leukaemia (AML)

LC–MS based global metabolite profiling: the necessity of high data quality

Contact Info

Product

Resources

About