Sparse PLS discriminant analysis: biologically relevant feature selection and graphical displays for multiclass problems

Key role of lipid management in nitrogen and aroma metabolism in an evolved wine yeast strain

Background

Fermentative aromas play a key role in the organoleptic profile of young wines. Their production depends both on yeast strain and fermentation conditions. A present-day trend in the wine industry consists in developing new strains with aromatic properties using adaptive evolution approaches. An evolved strain, Affinity™ ECA5, overproducing esters, was recently obtained. In this study, dynamics of nitrogen consumption and of the fermentative aroma synthesis of the evolved and its ancestral strains were compared and coupled with a transcriptomic analysis approach to better understand the metabolic reshaping of Affinity™ ECA5.

Results

Nitrogen assimilation was different between the two strains, particularly amino acids transported by carriers regulated by nitrogen catabolite repression. We also observed differences in the kinetics of fermentative aroma production, especially in the bioconversion of higher alcohols into acetate esters. Finally, transcriptomic data showed that the enhanced bioconversion into acetate esters by the evolved strain was associated with the repression of genes involved in sterol biosynthesis rather than an enhanced expression of ATF1 and ATF2 (genes coding for the enzymes responsible for the synthesis of acetate esters from higher alcohols).

Conclusions

An integrated approach to yeast metabolism—combining transcriptomic analyses and online monitoring data—showed differences between the two strains at different levels. Differences in nitrogen source consumption were observed suggesting modifications of NCR in the evolved strain. Moreover, the evolved strain showed a different way of managing the lipid source, which notably affected the production of acetate esters, likely because of a greater availability of acetyl-CoA for the evolved strain.

Electronic supplementary material

The online version of this article (doi:10.1186/s12934-016-0434-6) contains supplementary material, which is available to authorized users.

Visualization and Differential Analysis of Protein Expression Data Using R

Data analysis is essential to derive meaningful conclusions from proteomic data. This chapter describes ways of performing common data visualization and differential analysis tasks on gel-based proteomic datasets using a freely available statistical software package (R). A workflow followed is illustrated using a synthetic dataset as example.

Chemometrics for ion mobility spectrometry data: recent advances and future prospects

This is the first comprehensive review on chemometric techniques used in ion mobility spectrometry data analysis.

Exploring transcriptomic diversity in muscle revealed that cellular signaling pathways mainly differentiate five Western porcine breeds

Background

Among transcriptomic studies, those comparing species or populations can increase our understanding of the impact of the evolutionary forces on the differentiation of populations. A particular situation is the one of short evolution time with breeds of a domesticated species that underwent strong selective pressures. In this study, the gene expression diversity across five pig breeds has been explored in muscle. Samples came from: 24 Duroc, 33 Landrace, 41 Large White dam line, 10 Large White sire line and 39 Piétrain. From these animals, 147 muscle samples obtained at slaughter were analyzed using the porcine Agilent 44 K v1 microarray.

Results

A total of 12,358 genes were identified as expressed in muscle after normalization and 1,703 genes were declared differential for at least one breed (FDR < 0.001). The functional analysis highlighted that gene expression diversity is mainly linked to cellular signaling pathways such as the PI3K (phosphoinositide 3-kinase) pathway. The PI3K pathway is known to be involved in the control of development of the skeletal muscle mass by affecting extracellular matrix - receptor interactions, regulation of actin cytoskeleton pathways and some metabolic functions. This study also highlighted 228 spots (171 unique genes) that differentiate the breeds from each other. A common subgroup of 15 genes selected by three statistical methods was able to differentiate Duroc, Large White and Piétrain breeds.

Conclusions

This study on transcriptomic differentiation across Western pig breeds highlighted a global picture: mainly signaling pathways were affected. This result is consistent with the selection objective of increasing muscle mass. These transcriptional changes may indicate selection pressure or simply breed differences which may be driven by human selection. Further work aiming at comparing genetic and transcriptomic diversities would further increase our understanding of the consequences of human impact on livestock species.

Electronic supplementary material

The online version of this article (doi:10.1186/s12864-015-2259-9) contains supplementary material, which is available to authorized users.

Serum Glycoprotein Biomarker Discovery and Qualification Pipeline Reveals Novel Diagnostic Biomarker Candidates for Esophageal Adenocarcinoma

We report an integrated pipeline for efficient serum glycoprotein biomarker candidate discovery and qualification that may be used to facilitate cancer diagnosis and management. The discovery phase used semi-automated lectin magnetic bead array (LeMBA)-coupled tandem mass spectrometry with a dedicated data-housing and analysis pipeline; GlycoSelector (http://glycoselector.di.uq.edu.au). The qualification phase used lectin magnetic bead array-multiple reaction monitoring-mass spectrometry incorporating an interactive web-interface, Shiny mixOmics (http://mixomics-projects.di.uq.edu.au/Shiny), for univariate and multivariate statistical analysis. Relative quantitation was performed by referencing to a spiked-in glycoprotein, chicken ovalbumin. We applied this workflow to identify diagnostic biomarkers for esophageal adenocarcinoma (EAC), a life threatening malignancy with poor prognosis in the advanced setting. EAC develops from metaplastic condition Barrett's esophagus (BE). Currently diagnosis and monitoring of at-risk patients is through endoscopy and biopsy, which is expensive and requires hospital admission. Hence there is a clinical need for a noninvasive diagnostic biomarker of EAC. In total 89 patient samples from healthy controls, and patients with BE or EAC were screened in discovery and qualification stages. Of the 246 glycoforms measured in the qualification stage, 40 glycoforms (as measured by lectin affinity) qualified as candidate serum markers. The top candidate for distinguishing healthy from BE patients' group was Narcissus pseudonarcissus lectin (NPL)-reactive Apolipoprotein B-100 (p value = 0.0231; AUROC = 0.71); BE versus EAC, Aleuria aurantia lectin (AAL)-reactive complement component C9 (p value = 0.0001; AUROC = 0.85); healthy versus EAC, Erythroagglutinin Phaseolus vulgaris (EPHA)-reactive gelsolin (p value = 0.0014; AUROC = 0.80). A panel of 8 glycoforms showed an improved AUROC of 0.94 to discriminate EAC from BE. Two biomarker candidates were independently verified by lectin magnetic bead array-immunoblotting, confirming the validity of the relative quantitation approach. Thus, we have identified candidate biomarkers, which, following large-scale clinical evaluation, can be developed into diagnostic blood tests. A key feature of the pipeline is the potential for rapid translation of the candidate biomarkers to lectin-immunoassays.

Meta-Analysis of Large-Scale Toxicogenomic Data Finds Neuronal Regeneration Related Protein and Cathepsin D to Be Novel Biomarkers of Drug-Induced Toxicity

Undesirable toxicity is one of the main reasons for withdrawing drugs from the market or eliminating them as candidates in clinical trials. Although numerous studies have attempted to identify biomarkers capable of predicting pharmacotoxicity, few have attempted to discover robust biomarkers that are coherent across various species and experimental settings. To identify such biomarkers, we conducted meta-analyses of massive gene expression profiles for 6,567 in vivo rat samples and 453 compounds. After applying rigorous feature reduction procedures, our analyses identified 18 genes to be related with toxicity upon comparisons of untreated versus treated and innocuous versus toxic specimens of kidney, liver and heart tissue. We then independently validated these genes in human cell lines. In doing so, we found several of these genes to be coherently regulated in both in vivo rat specimens and in human cell lines. Specifically, mRNA expression of neuronal regeneration-related protein was robustly down-regulated in both liver and kidney cells, while mRNA expression of cathepsin D was commonly up-regulated in liver cells after exposure to toxic concentrations of chemical compounds. Use of these novel toxicity biomarkers may enhance the efficiency of screening for safe lead compounds in early-phase drug development prior to animal testing.

Gene Function Prediction from Functional Association Networks Using Kernel Partial Least Squares Regression

With the growing availability of large-scale biological datasets, automated methods of extracting functionally meaningful information from this data are becoming increasingly important. Data relating to functional association between genes or proteins, such as co-expression or functional association, is often represented in terms of gene or protein networks. Several methods of predicting gene function from these networks have been proposed. However, evaluating the relative performance of these algorithms may not be trivial: concerns have been raised over biases in different benchmarking methods and datasets, particularly relating to non-independence of functional association data and test data. In this paper we propose a new network-based gene function prediction algorithm using a commute-time kernel and partial least squares regression (Compass). We compare Compass to GeneMANIA, a leading network-based prediction algorithm, using a number of different benchmarks, and find that Compass outperforms GeneMANIA on these benchmarks. We also explicitly explore problems associated with the non-independence of functional association data and test data. We find that a benchmark based on the Gene Ontology database, which, directly or indirectly, incorporates information from other databases, may considerably overestimate the performance of algorithms exploiting functional association data for prediction.

The quest for tolerant varieties: the importance of integrating "omics" techniques to phenotyping

The primary objective of crop breeding is to improve yield and/or harvest quality while minimizing inputs. Global climate change and the increase in world population are significant challenges for agriculture and call for further improvements to crops and the development of new tools for research. Significant progress has been made in the molecular and genetic analysis of model plants. However, is science generating false expectations? Are 'omic techniques generating valuable information that can be translated into the field? The exploration of crop biodiversity and the correlation of cellular responses to stress tolerance at the plant level is currently a challenge. This viewpoint reviews concisely the problems one encounters when working on a crop and provides an outline of possible workflows when initiating cellular phenotyping via "-omic" techniques (transcriptomics, proteomics, metabolomics).

RPCA-Based Tumor Classification Using Gene Expression Data

Microarray techniques have been used to delineate cancer groups or to identify candidate genes for cancer prognosis. As such problems can be viewed as classification ones, various classification methods have been applied to analyze or interpret gene expression data. In this paper, we propose a novel method based on robust principal component analysis (RPCA) to classify tumor samples of gene expression data. Firstly, RPCA is utilized to highlight the characteristic genes associated with a special biological process. Then, RPCA and RPCA+LDA (robust principal component analysis and linear discriminant analysis) are used to identify the features. Finally, support vector machine (SVM) is applied to classify the tumor samples of gene expression data based on the identified features. Experiments on seven data sets demonstrate that our methods are effective and feasible for tumor classification.

Development of a Drug-Response Modeling Framework to Identify Cell Line Derived Translational Biomarkers That Can Predict Treatment Outcome to Erlotinib or Sorafenib

Development of drug responsive biomarkers from pre-clinical data is a critical step in drug discovery, as it enables patient stratification in clinical trial design. Such translational biomarkers can be validated in early clinical trial phases and utilized as a patient inclusion parameter in later stage trials. Here we present a study on building accurate and selective drug sensitivity models for Erlotinib or Sorafenib from pre-clinical in vitro data, followed by validation of individual models on corresponding treatment arms from patient data generated in the BATTLE clinical trial. A Partial Least Squares Regression (PLSR) based modeling framework was designed and implemented, using a special splitting strategy and canonical pathways to capture robust information for model building. Erlotinib and Sorafenib predictive models could be used to identify a sub-group of patients that respond better to the corresponding treatment, and these models are specific to the corresponding drugs. The model derived signature genes reflect each drug’s known mechanism of action. Also, the models predict each drug’s potential cancer indications consistent with clinical trial results from a selection of globally normalized GEO expression datasets.

Lycopene bioavailability is associated with a combination of genetic variants

The intake of tomatoes and tomato products, which constitute the main dietary source of the red pigment lycopene (LYC), has been associated with a reduced risk of prostate cancer and cardiovascular disease, suggesting a protective role of this carotenoid. However, LYC bioavailability displays high interindividual variability. This variability may lead to varying biological effects following LYC consumption. Based on recent results obtained with two other carotenoids, we assumed that this variability was due, at least in part, to several single nucleotide polymorphisms (SNPs) in genes involved in LYC and lipid metabolism. Thus, we aimed at identifying a combination of SNPs significantly associated with the variability in LYC bioavailability. In a postprandial study, 33 healthy male volunteers consumed a test meal containing 100g tomato puree, which provided 9.7 mg all-trans LYC. LYC concentrations were measured in plasma chylomicrons (CM) isolated at regular time intervals over 8 h postprandially. For the study 1885 SNPs in 49 candidate genes, i.e., genes assumed to play a role in LYC bioavailability, were selected. Multivariate statistical analysis (partial least squares regression) was used to identify and validate the combination of SNPs most closely associated with postprandial CM LYC response. The postprandial CM LYC response to the meal was notably variable with a CV of 70%. A significant (P=0.037) and validated partial least squares regression model, which included 28 SNPs in 16 genes, explained 72% of the variance in the postprandial CM LYC response. The postprandial CM LYC response was also positively correlated to fasting plasma LYC concentrations (r=0.37, P<0.05). The ability to respond to LYC is explained, at least partly, by a combination of 28 SNPs in 16 genes. Interindividual variability in bioavailability apparently affects the long-term blood LYC status, which could ultimately modulate the biological response following LYC supplementation.

Early-life establishment of the swine gut microbiome and impact on host phenotypes

Early bacterial colonization and succession within the gastrointestinal tract has been suggested to be crucial in the establishment of specific microbiota composition and the shaping of host phenotype. Here, the composition and dynamics of faecal microbiomes were studied for 31 healthy piglets across five age strata (days 14, 36, 48, 60 and 70 after birth) together with their mothers. Faecal microbiome composition was assessed by 16S rRNA gene 454-pyrosequencing. Bacteroidetes and Firmicutes were the predominant phyla present at each age. For all piglets, luminal secretory IgA concentration was measured at day 70, and body weight was recorded until day 70. The microbiota of suckling piglets was mainly represented by Bacteroides, Oscillibacter, Escherichia/Shigella, Lactobacillus and unclassified Ruminococcaceae genera. This pattern contrasted with that of Acetivibrio, Dialister, Oribacterium, Succinivibrio and Prevotella genera, which appeared increased after weaning. Lactobacillus fermentum might be vertically transferred via breast milk or faeces. The microbiota composition coevolved with their hosts towards two different clusters after weaning, primarily distinguished by unclassified Ruminococcaceae and Prevotella abundances. Prevotella was positively correlated with luminal secretory IgA concentrations, and body weight. Our study opens up new possibilities for health and feed efficiency manipulation via genetic selection and nutrition in the agricultural domain.

Temperament type specific metabolite profiles of the prefrontal cortex and serum in cattle

In the past decade the number of studies investigating temperament in farm animals has increased greatly because temperament has been shown not only to affect handling but also reproduction, health and economically important production traits. However, molecular pathways underlying temperament and molecular pathways linking temperament to production traits, health and reproduction have yet to be studied in full detail. Here we report the results of metabolite profiling of the prefrontal cortex and serum of cattle with distinct temperament types that were performed to further explore their molecular divergence in the response to the slaughter procedure and to identify new targets for further research of cattle temperament. By performing an untargeted comprehensive metabolite profiling, 627 and 1097 metabolite features comprising 235 and 328 metabolites could be detected in the prefrontal cortex and serum, respectively. In total, 54 prefrontal cortex and 51 serum metabolite features were indicated to have a high relevance in the classification of temperament types by a sparse partial least square discriminant analysis. A clear discrimination between fearful/neophobic-alert, interested-stressed, subdued/uninterested-calm and outgoing/neophilic-alert temperament types could be observed based on the abundance of the identified relevant prefrontal cortex and serum metabolites. Metabolites with high relevance in the classification of temperament types revealed that the main differences between temperament types in the response to the slaughter procedure were related to the abundance of glycerophospholipids, fatty acyls and sterol lipids. Differences in the abundance of metabolites related to C21 steroid metabolism and oxidative stress indicated that the differences in the metabolite profiles of the four extreme temperament types could be the result of a temperament type specific regulation of molecular pathways that are known to be involved in the stress and fear response.

Improving PLS-RFE based gene selection for microarray data classification

Gene selection plays a crucial role in constructing efficient classifiers for microarray data classification, since microarray data is characterized by high dimensionality and small sample sizes and contains irrelevant and redundant genes. In practical use, partial least squares-based gene selection approaches can obtain gene subsets of good qualities, but are considerably time-consuming. In this paper, we propose to integrate partial least squares based recursive feature elimination (PLS-RFE) with two feature elimination schemes: simulated annealing and square root, respectively, to speed up the feature selection process. Inspired from the strategy of annealing schedule, the two proposed approaches eliminate a number of features rather than one least informative feature during each iteration and the number of removed features decreases as the iteration proceeds. To verify the effectiveness and efficiency of the proposed approaches, we perform extensive experiments on six publicly available microarray data with three typical classifiers, including Naïve Bayes, K-Nearest-Neighbor and Support Vector Machine, and compare our approaches with ReliefF, PLS and PLS-RFE feature selectors in terms of classification accuracy and running time. Experimental results demonstrate that the two proposed approaches accelerate the feature selection process impressively without degrading the classification accuracy and obtain more compact feature subsets for both two-category and multi-category problems. Further experimental comparisons in feature subset consistency show that the proposed approach with simulated annealing scheme not only has better time performance, but also obtains slightly better feature subset consistency than the one with square root scheme.

Intestinal microbiota-derived metabolomic blood plasma markers for prior radiation injury

PURPOSE

Assessing whole-body radiation injury and absorbed dose is essential for remediation efforts following accidental or deliberate exposure in medical, industrial, military, or terrorist incidents. We hypothesize that variations in specific metabolite concentrations extracted from blood plasma would correlate with whole-body radiation injury and dose.

METHODS AND MATERIALS

Groups of C57BL/6 mice (n=12 per group) were exposed to 0, 2, 4, 8, and 10.4 Gy of whole-body gamma radiation. At 24 hours after treatment, all animals were euthanized, and both plasma and liver biopsy samples were obtained, the latter being used to identify a distinct hepatic radiation injury response within plasma. A semiquantitative, untargeted metabolite/lipid profile was developed using gas chromatography-mass spectrometry and liquid chromatography-tandem mass spectrometry, which identified 354 biochemical compounds. A second set of C57BL/6 mice (n=6 per group) were used to assess a subset of identified plasma markers beyond 24 hours.

RESULTS

We identified a cohort of 37 biochemical compounds in plasma that yielded the optimal separation of the irradiated sample groups, with the most correlated metabolites associated with pyrimidine (positively correlated) and tryptophan (negatively correlated) metabolism. The latter were predominantly associated with indole compounds, and there was evidence that these were also correlated between liver and plasma. No evidence of saturation as a function of dose was observed, as has been noted for studies involving metabolite analysis of urine.

CONCLUSIONS

Plasma profiling of specific metabolites related to pyrimidine and tryptophan pathways can be used to differentiate whole-body radiation injury and dose response. As the tryptophan-associated indole compounds have their origin in the intestinal microbiome and subsequently the liver, these metabolites particularly represent an attractive marker for radiation injury within blood plasma.

CMV reactivation drives posttransplant T-cell reconstitution and results in defects in the underlying TCRβ repertoire

Although cytomegalovirus (CMV) reactivation has long been implicated in posttransplant immune dysfunction, the molecular mechanisms that drive this phenomenon remain undetermined. To address this, we combined multiparameter flow cytometric analysis and T-cell subpopulation sorting with high-throughput sequencing of the T-cell repertoire, to produce a thorough evaluation of the impact of CMV reactivation on T-cell reconstitution after unrelated-donor hematopoietic stem cell transplant. We observed that CMV reactivation drove a >50-fold specific expansion of Granzyme B(high)/CD28(low)/CD57(high)/CD8(+) effector memory T cells (Tem) and resulted in a linked contraction of all naive T cells, including CD31(+)/CD4(+) putative thymic emigrants. T-cell receptor β (TCRβ) deep sequencing revealed a striking contraction of CD8(+) Tem diversity due to CMV-specific clonal expansions in reactivating patients. In addition to querying the topography of the expanding CMV-specific T-cell clones, deep sequencing allowed us, for the first time, to exhaustively evaluate the underlying TCR repertoire. Our results reveal new evidence for significant defects in the underlying CD8 Tem TCR repertoire in patients who reactivate CMV, providing the first molecular evidence that, in addition to driving expansion of virus-specific cells, CMV reactivation has a detrimental impact on the integrity and heterogeneity of the rest of the T-cell repertoire. This trial was registered at www.clinicaltrials.gov as #NCT01012492.

Fourteen years of plant proteomics reflected in Proteomics: moving from model species and 2DE-based approaches to orphan species and gel-free platforms

In this article, the topic of plant proteomics is reviewed based on related papers published in the journal Proteomics since publication of the first issue in 2001. In total, around 300 original papers and 41 reviews published in Proteomics between 2000 and 2014 have been surveyed. Our main objective for this review is to help bridge the gap between plant biologists and proteomics technologists, two often very separate groups. Over the past years a number of reviews on plant proteomics have been published . To avoid repetition we have focused on more recent literature published after 2010, and have chosen to rather make continuous reference to older publications. The use of the latest proteomics techniques and their integration with other approaches in the "systems biology" direction are discussed more in detail. Finally we comment on the recent history, state of the art, and future directions of plant proteomics, using publications in Proteomics to illustrate the progress in the field. The review is organized into two major blocks, the first devoted to provide an overview of experimental systems (plants, plant organs, biological processes) and the second one to the methodology.

Predicting cytotoxicity of complex mixtures in high cancer incidence regions of the Huai River Basin based on GC-MS spectrum with partial least squares regression

Complex mixture exposures, such as those associated with water sources, are an important issue in health risk assessment. This study assessed the cytotoxicity of chemical mixtures extracted from water sources in regions of the Huai River Basin with high cancer incidences and built statistical models of cytotoxicity based on pollution profiles that were measured with gas chromatography-mass spectrometry (GC-MS). Both surface and ground waters were collected from rural water sources of Shenqiu County, Henan Province of China from 2008 to 2011 and extracted with XAD-2 resigns. Cytotoxicity was evaluated with Chinese hamster ovary K1 (CHO-K1) cells and compared against the pollution profiles of the extracts. IC50 of water samples ranged from 0.023 to 0.338L-eq/mL. The pollutants in waters determined by GC-MS are complex and some of the compounds that contributed to cytotoxicity lack toxicity data. A partial least squares (PLS) regression model of cytotoxicity was built based on linear aggregation of predictor variables (i.e., peaks for single compounds in the gas chromatograms). The PLS model contains 2 PLS factors extracted from 141 variables. The model was validated internally with training data permutation and externally with a test sample. The model explained 92% of the cytotoxicity in the training samples and 40% in the test sample. This approach provides a general, rapid method for relating water toxicity to GC-MS chromatograms and for predicting the compounds that contribute most to toxicity.

Patterns of brain structural connectivity differentiate normal weight from overweight subjects

Background

Alterations in the hedonic component of ingestive behaviors have been implicated as a possible risk factor in the pathophysiology of overweight and obese individuals. Neuroimaging evidence from individuals with increasing body mass index suggests structural, functional, and neurochemical alterations in the extended reward network and associated networks.

Aim

To apply a multivariate pattern analysis to distinguish normal weight and overweight subjects based on gray and white-matter measurements.

Methods

Structural images (N = 120, overweight N = 63) and diffusion tensor images (DTI) (N = 60, overweight N = 30) were obtained from healthy control subjects. For the total sample the mean age for the overweight group (females = 32, males = 31) was 28.77 years (SD = 9.76) and for the normal weight group (females = 32, males = 25) was 27.13 years (SD = 9.62). Regional segmentation and parcellation of the brain images was performed using Freesurfer. Deterministic tractography was performed to measure the normalized fiber density between regions. A multivariate pattern analysis approach was used to examine whether brain measures can distinguish overweight from normal weight individuals.

Results

1. White-matter classification: The classification algorithm, based on 2 signatures with 17 regional connections, achieved 97% accuracy in discriminating overweight individuals from normal weight individuals. For both brain signatures, greater connectivity as indexed by increased fiber density was observed in overweight compared to normal weight between the reward network regions and regions of the executive control, emotional arousal, and somatosensory networks. In contrast, the opposite pattern (decreased fiber density) was found between ventromedial prefrontal cortex and the anterior insula, and between thalamus and executive control network regions. 2. Gray-matter classification: The classification algorithm, based on 2 signatures with 42 morphological features, achieved 69% accuracy in discriminating overweight from normal weight. In both brain signatures regions of the reward, salience, executive control and emotional arousal networks were associated with lower morphological values in overweight individuals compared to normal weight individuals, while the opposite pattern was seen for regions of the somatosensory network.

Conclusions

1. An increased BMI (i.e., overweight subjects) is associated with distinct changes in gray-matter and fiber density of the brain. 2. Classification algorithms based on white-matter connectivity involving regions of the reward and associated networks can identify specific targets for mechanistic studies and future drug development aimed at abnormal ingestive behavior and in overweight/obesity.

•

Multivariate analysis can be used to classify overweight from normal weight individuals.

•

Anatomical connectivity achieved 97% accuracy in the classification algorithm.

•

Greater connectivity was observed in extended reward and somatosensory regions.

•

Morphological gray-matter achieved 69% accuracy in the classification algorithm.

•

Lower morphological values were observed in regions of the extended reward network.

Linking metabolomics data to underlying metabolic regulation

The comprehensive experimental analysis of a metabolic constitution plays a central role in approaches of organismal systems biology. Quantifying the impact of a changing environment on the homeostasis of cellular metabolism has been the focus of numerous studies applying various metabolomics techniques. It has been proven that approaches which integrate different analytical techniques, e.g., LC-MS, GC-MS, CE-MS and H-NMR, can provide a comprehensive picture of a certain metabolic homeostasis. Identification of metabolic compounds and quantification of metabolite levels represent the groundwork for the analysis of regulatory strategies in cellular metabolism. This significantly promotes our current understanding of the molecular organization and regulation of cells, tissues and whole organisms. Nevertheless, it is demanding to elicit the pertinent information which is contained in metabolomics data sets. Based on the central dogma of molecular biology, metabolite levels and their fluctuations are the result of a directed flux of information from gene activation over transcription to translation and posttranslational modification. Hence, metabolomics data represent the summed output of a metabolic system comprising various levels of molecular organization. As a consequence, the inverse assignment of metabolomics data to underlying regulatory processes should yield information which—if deciphered correctly—provides comprehensive insight into a metabolic system. Yet, the deduction of regulatory principles is complex not only due to the high number of metabolic compounds, but also because of a high level of cellular compartmentalization and differentiation. Motivated by the question how metabolomics approaches can provide a representative view on regulatory biochemical processes, this article intends to present and discuss current metabolomics applications, strategies of data analysis and their limitations with respect to the interpretability in context of biological processes.

Novel multivariate methods for integration of genomics and proteomics data: applications in a kidney transplant rejection study

Multi-omics research is a key ingredient of data-intensive life sciences research, permitting measurement of biological molecules at different functional levels in the same individual. For a complete picture at the biological systems level, appropriate statistical techniques must however be developed to integrate different 'omics' data sets (e.g., genomics and proteomics). We report here multivariate projection-based analyses approaches to genomics and proteomics data sets, using the case study of and applications to observations in kidney transplant patients who experienced an acute rejection event (n=20) versus non-rejecting controls (n=20). In this data sets, we show how these novel methodologies might serve as promising tools for dimension reduction and selection of relevant features for different analytical frameworks. Unsupervised analyses highlighted the importance of post transplant time-of-rejection, while supervised analyses identified gene and protein signatures that together predicted rejection status with little time effect. The selected genes are part of biological pathways that are representative of immune responses. Gene enrichment profiles revealed increases in innate immune responses and neutrophil activities and a depletion of T lymphocyte related processes in rejection samples as compared to controls. In all, this article offers candidate biomarkers for future detection and monitoring of acute kidney transplant rejection, as well as ways forward for methodological advances to better harness multi-omics data sets.

Metabolomics profile comparisons of irradiated and nonirradiated stored donor red blood cells

BACKGROUND

Understanding the metabolites that are altered by donor red blood cell (RBC) storage and irradiation may provide insight into the metabolic pathways disrupted by the RBC storage lesion.

STUDY DESIGN AND METHODS

Patterns of metabolites, representing more than 11,000 distinct mass-to-charge ratio (m/z) features, were compared between gamma-irradiated and nonirradiated CPDA-1-split RBCs from six human donors over 35 days of storage using multilevel sparse partial least squares discriminant analysis (msPLSDA), hierarchical clustering, pathway enrichment analysis, and network analysis.

RESULTS

In msPLSDA analysis, RBC units stored 7 days or fewer (irradiated or nonirradiated) showed similar metabolomic profiles. By contrast, donor RBCs stored 10 days or more demonstrated distinct clustering as a function of storage time and irradiation. Irradiation shifted metabolic features to those seen in older units. Hierarchical clustering analysis identified at least two clusters of metabolites that differentiated between RBC units based on storage time and irradiation exposure, confirming results of the msPLSDA analysis. Pathway enrichment analysis, used to map the discriminatory biochemical features to specific metabolic pathways, identified four pathways significantly affected by irradiation and/or storage including arachidonic acid (p = 3.3 × 10(-33)) and linoleic acid (p = 1.61 × 10(-11)) metabolism.

CONCLUSION

RBC storage under blood bank conditions produces numerous metabolic alterations. Gamma irradiation accentuates these differences as the age of blood increases, indicating that at the biochemical level irradiation accelerates metabolic aging of stored RBCs. Metabolites involved in the cellular membrane are prominently affected and may be useful biomarkers of the RBC storage lesion.

Current advances in systems and integrative biology

Systems biology has gained a tremendous amount of interest in the last few years. This is partly due to the realization that traditional approaches focusing only on a few molecules at a time cannot describe the impact of aberrant or modulated molecular environments across a whole system. Furthermore, a hypothesis-driven study aims to prove or disprove its postulations, whereas a hypothesis-free systems approach can yield an unbiased and novel testable hypothesis as an end-result. This latter approach foregoes assumptions which predict how a biological system should react to an altered microenvironment within a cellular context, across a tissue or impacting on distant organs. Additionally, re-use of existing data by systematic data mining and re-stratification, one of the cornerstones of integrative systems biology, is also gaining attention. While tremendous efforts using a systems methodology have already yielded excellent results, it is apparent that a lack of suitable analytic tools and purpose-built databases poses a major bottleneck in applying a systematic workflow. This review addresses the current approaches used in systems analysis and obstacles often encountered in large-scale data analysis and integration which tend to go unnoticed, but have a direct impact on the final outcome of a systems approach. Its wide applicability, ranging from basic research, disease descriptors, pharmacological studies, to personalized medicine, makes this emerging approach well suited to address biological and medical questions where conventional methods are not ideal.

Interindividual variability of lutein bioavailability in healthy men: characterization, genetic variants involved, and relation with fasting plasma lutein concentration

BACKGROUND

Lutein accumulates in the macula and brain, where it is assumed to play physiologic roles. The bioavailability of lutein is assumed to display a high interindividual variability that has been hypothesized to be attributable, at least partly, to genetic polymorphisms.

OBJECTIVES

We characterized the interindividual variability in lutein bioavailability in humans, assessed the relation between this variability and the fasting blood lutein concentration, and identified single nucleotide polymorphisms (SNPs) involved in this phenomenon.

DESIGN

In a randomized, 2-way crossover study, 39 healthy men consumed a meal that contained a lutein supplement or the same meal for which lutein was provided through a tomato puree. The lutein concentration was measured in plasma chylomicrons isolated at regular time intervals over 8 h postprandially. Multivariate statistical analyses were used to identify a combination of SNPs associated with the postprandial chylomicron lutein response (0-8-h area under the curve). A total of 1785 SNPs in 51 candidate genes were selected.

RESULTS

Postprandial chylomicron lutein responses to meals were very variable (CV of 75% and 137% for the lutein-supplement meal and the meal with tomato-sourced lutein, respectively). Postprandial chylomicron lutein responses measured after the 2 meals were positively correlated (r = 0.68, P < 0.0001) and positively correlated to the fasting plasma lutein concentration (r = 0.51, P < 0.005 for the lutein-supplement-containing meal). A significant (P = 1.9 × 10(-4)) and validated partial least-squares regression model, which included 29 SNPs in 15 genes, explained most of the variance in the postprandial chylomicron lutein response.

CONCLUSIONS

The ability to respond to lutein appears to be, at least in part, genetically determined. The ability is explained, in large part, by a combination of SNPs in 15 genes related to both lutein and chylomicron metabolism. Finally, our results suggest that the ability to respond to lutein and blood lutein status are related. This trial was registered at clinicaltrials.gov as NCT02100774.

The spatiotemporal program of DNA replication is associated with specific combinations of chromatin marks in human cells

The duplication of mammalian genomes is under the control of a spatiotemporal program that orchestrates the positioning and the timing of firing of replication origins. The molecular mechanisms coordinating the activation of about predicted origins remain poorly understood, partly due to the intrinsic rarity of replication bubbles, making it difficult to purify short nascent strands (SNS). The precise identification of origins based on the high-throughput sequencing of SNS constitutes a new methodological challenge. We propose a new statistical method with a controlled resolution, adapted to the detection of replication origins from SNS data. We detected an average of 80,000 replication origins in different cell lines. To evaluate the consistency between different protocols, we compared SNS detections with bubble trapping detections. This comparison demonstrated a good agreement between genome-wide methods, with 65% of SNS-detected origins validated by bubble trapping, and 44% of bubble trapping origins validated by SNS origins, when compared at the same resolution. We investigated the interplay between the spatial and the temporal programs of replication at fine scales. We show that most of the origins detected in regions replicated in early S phase are shared by all the cell lines investigated whereas cell-type-specific origins tend to be replicated in late S phase. We shed a new light on the key role of CpG islands, by showing that 80% of the origins associated with CGIs are constitutive. Our results further show that at least 76% of CGIs are origins of replication. The analysis of associations with chromatin marks at different timing of cell division revealed new potential epigenetic regulators driving the spatiotemporal activity of replication origins. We highlight the potential role of H4K20me1 and H3K27me3, the coupling of which is correlated with increased efficiency of replication origins, clearly identifying those marks as potential key regulators of replication origins.

Replication is the mechanism by which genomes are duplicated into two exact copies. Genomic stability is under the control of a spatiotemporal program that orchestrates both the positioning and the timing of firing of about 50,000 replication starting points, also called replication origins. Replication bubbles found at origins have been very difficult to map due to their short lifespan. Moreover, with the flood of data characterizing new sequencing technologies, the precise statistical analysis of replication data has become an additional challenge. We propose a new method to map replication origins on the human genome, and we assess the reliability of our finding using experimental validation and comparison with origins maps obtained by bubble trapping. This fine mapping then allowed us to identify potential regulators of the replication dynamics. Our study highlights the key role of CpG Islands and identifies new potential epigenetic regulators (methylation of lysine 4 on histone H4, and tri-methylation of lysine 27 on histone H3) whose coupling is correlated with an increase in the efficiency of replication origins, suggesting those marks as potential key regulators of replication. Overall, our study defines new potentially important pathways that might regulate the sequential firing of origins during genome duplication.

Floral scent in natural hybrids of Ipomopsis (Polemoniaceae) and their parental species

BACKGROUND AND AIMS

Floral traits, such as floral volatiles, can contribute to pre-zygotic reproductive isolation by promoting species-specific pollinator foraging. When hybrid zones form, floral traits could also influence post-zygotic isolation. This study examined floral volatiles in parental species and natural hybrids in order to explore potential scent mediation of pre-zygotic and post-zygotic isolation.

METHODS

Floral bouquets were analysed for the sister species Ipomopsis aggregata and I. tenuituba and their natural hybrids at two contact sites differing in both hybridization rate and temporal foraging pattern of hawkmoth pollinators. Floral volatiles were quantified in diurnal and nocturnal scent samples using gas chromatography-mass spectrometry.

KEY RESULTS

The bouquets of parental species and hybrids showed qualitative overlap. All flowers emitted similar sets of monoterpenoid, sesquiterpenoid, aliphatic and benzenoid compounds, but separated into groups defined by multivariate analysis of quantitative emissions. The parental species differed most strikingly in the nitrogenous compound indole, which was found almost exclusively in nocturnal bouquets of I. tenuituba. Natural hybrid bouquets were highly variable, and showed emission rates of several compounds that appeared transgressive. However, indole emission rates were intermediate in the hybrids compared with rates in the parents. Volatile bouquets at the contact site with lower hybridization did not show greater species specificity in overall scent emission, but I. tenuituba presented a stronger indole signal during peak hawkmoth activity at that site.

CONCLUSIONS

The two species of Ipomopsis differed in patterns of floral bouquets, with indole emitted in nocturnal I. tenuituba, but not in I. aggregata. Natural hybrid bouquets were not consistently intermediate between the parents, although hybrids were intermediate in indole emission. The indole signal could potentially serve as a hawkmoth attractant that mediates reproductive isolation both before and after hybrid formation.

Impairment of adipose tissue in Prader-Willi syndrome rescued by growth hormone treatment

BACKGROUND

Prader-Willi syndrome (PWS) results from abnormalities in the genomic imprinting process leading to hypothalamic dysfunction with an alteration of growth hormone (GH) secretion. PWS is associated with early morbid obesity and short stature which can be efficiently improved with GH treatment.

OBJECTIVES

Our aims were to highlight adipose tissue structural and functional impairments in children with PWS and to study the modifications of those parameters on GH treatment.

SUBJECTS AND METHODS

Plasma samples and adipose tissue biopsies were obtained from 23 research centers in France coordinated by the reference center for PWS in Toulouse, France. Lean controls (n=33), non-syndromic obese (n=53), untreated (n=26) and GH-treated PWS (n=43) children were enrolled in the study. Adipose tissue biopsies were obtained during scheduled surgeries from 15 lean control, 7 untreated and 8 GH-treated PWS children.

RESULTS

Children with PWS displayed higher insulin sensitivity as shown by reduced glycemia, insulinemia and HOMA-IR compared with non-syndromic obese children. In contrast, plasma inflammatory cytokines such as TNF-α, MCP-1 and IL-8 were increased in PWS. Analysis of biopsies compared with control children revealed decreased progenitor cell content in the stromal vascular fraction of adipose tissue and an impairment of lipolytic response to β-adrenergic agonist in PWS adipocytes. Interestingly, both of these alterations in PWS seem to be ameliorated on GH treatment.

CONCLUSION

Herein, we report adipose tissue dysfunctions in children with PWS which may be partially restored by GH treatment.

Standardization of data processing and statistical analysis in comparative plant proteomics experiment

Two-dimensional gel electrophoresis remains the most widely used technique for protein separation in plant proteomics experiments. Despite the continuous technical advances and improvements in current 2-DE protocols, an adequate and correct experimental design and statistical analysis of the data tend to be ignored or not properly documented in current literature. Both proper experimental design and appropriate statistical analysis are requested in order to confidently discuss our results and to conclude from experimental data.In this chapter, we describe a model procedure for a correct experimental design and a complete statistical analysis of proteomic dataset. Our model procedure covers all of the steps in data mining and processing, starting with the data preprocessing (transformation, missing value imputation, definition of outliers) and univariate statistics (parametric and nonparametric tests), and finishing with multivariate statistics (clustering, heat-mapping, PCA, ICA, PLS-DA).

Sparse partial-least-squares discriminant analysis for different geographical origins of Salvia miltiorrhiza by (1) H-NMR-based metabolomics

INTRODUCTION

(1) H nuclear magnetic resonance (NMR) spectroscopy has clear advantages in respect of detecting various primary and secondary metabolites in plants simultaneously, non-targeted and non-destructively.

OBJECTIVE

To establish a method for detecting both primary and secondary metabolites in Salvia miltiorrhiza and screening potential geographical biomarkers effectively.

METHODS

Primary and secondary metabolites of S. militiorrhiza were detected and identified by (1) H-NMR fingerprint. Sparse partial-least-squares discriminant analysis (sPLS-DA) was undertaken for classification and variable selection in a one-step procedure and the classification error rates were implemented to estimate the cluster validation of sPLS-DA. Potential candidate metabolites by characterised different geographical origins of S. miltiorrhiza were identified according to the sparse loading vectors. The levels of these metabolites were quantified and evaluated by Kruskal-Wallis tests and also showed significant difference.

RESULTS

Twenty-six primary and secondary metabolites were identified in samples from different regions. The results suggest that malonate and succinate can be possibly recognised as the key markers for discriminating the geographical origin of S. miltiorrhiza based on the regulation and influence on the root respiratory rates of plants.

CONCLUSION

(1) H-NMR metabolic profiling combination with PLS-DA provided a very efficient and visualised representation of similarities and dissimilarities between S. miltiorrhiza samples.

The peripheral blood transcriptome reflects variations in immunity traits in swine: towards the identification of biomarkers

BACKGROUND

Immune traits (ITs) are potentially relevant criteria to characterize an individual's immune response. Our aim was to investigate whether the peripheral blood transcriptome can provide a significant and comprehensive view of IT variations in pig.

RESULTS

Sixty-day-old Large White pigs classified as extreme for in vitro production of IL2, IL10, IFNγ and TNFα, phagocytosis activity, in vivo CD4⁻/CD8⁺ or TCRγδ + cell counts, and anti-Mycoplasma antibody levels were chosen to perform a blood transcriptome analysis with a porcine generic array enriched with immunity-related genes. Differentially expressed (DE) genes for in vitro production of IL2 and IL10, phagocytosis activity and CD4⁻/CD8⁺ cell counts were identified. Gene set enrichment analysis revealed a significant over-representation of immune response functions. To validate the microarray-based results, a subset of DE genes was confirmed by RT-qPCR. An independent set of 74 animals was used to validate the covariation between gene expression levels and ITs. Five potential gene biomarkers were found for prediction of IL2 (RALGDS), phagocytosis (ALOX12) or CD4⁻/CD8⁺ cell count (GNLY, KLRG1 and CX3CR1). On average, these biomarkers performed with a sensitivity of 79% and a specificity of 86%.

CONCLUSIONS

Our results confirmed that gene expression profiling in blood represents a relevant molecular phenotype to refine ITs in pig and to identify potential biomarkers that can provide new insights into immune response analysis.

The intellectual disability of trisomy 21: differences in gene expression in a case series of patients with lower and higher IQ

Trisomy 21 (T21), or Down syndrome (DS), is the most frequent and recognizable cause of intellectual disabilities. The level of disability, as evaluated by the intelligence quotient (IQ) test, varies considerably between patients independent of other factors. To determine the genetic or molecular basis of this difference, a high throughput transcriptomic analysis was performed on twenty T21 patients with high and low IQ, and 10 healthy controls using Digital Gene Expression. More than 90 millions of tags were sequenced in the three libraries. A total of 80 genes of potential interest were selected for the qPCR experiment validation, and three housekeeping genes were used for normalizing purposes. HLA DQA1 and HLA DRB1 were significantly downregulated among the patients with a low IQ, the values found in the healthy controls being intermediate between those noted in the IQ+ and IQ- T21 patients. Interestingly, the intergenic region between these genes contains a binding sequence for the CCCTC-binding factor, or CTCF, and cohesin (a multisubunit complex), both of which are essential for expression of HLA DQA1 and HLA DRB1 and numerous other genes. Our results might lead to the discovery of genes, or genetic markers, that are directly involved in several phenotypes of DS and, eventually, to the identification of potential targets for therapeutic interventions.

MicroRNA signatures in hereditary breast cancer

This study aims to identify signatures of miR associated with hereditary, BRCA1 or BRCA2 mutation positive breast cancer (BC), and non-hereditary BC, either sporadic (SBC) or non-informative (BRCAX). Moreover, we search for signatures associated with tumor stage, immunohistochemistry and tumor molecular profile. Twenty formalin fixed paraffin embedded (FFPE) BCs, BRCA1, BRCA2, BRCAX and SBC, five per group were studied. Affymetrix platform miRNA v.3.0 was used to perform miR expression analysis. ER, PR, HER2 and Ki67 protein expression was analyzed by immunohistochemistry. BRCA1, BRCA2 and RASSF1 methylation analysis, AURKA copy number variations, and BRCA1 and BRCA2 deletions, were studied by MLPA. We validated eight of the miR selected by the arrays in 77 BCs by qRT-PCR. The miR profiles associated with tumor features were studied applying the Sparse Partial Least Squares Discriminant Analysis. MiR discrimination capability to distinguish hereditary and non-hereditary BC was analyzed by the discriminant function. With 15 out of 1,733 hsa-miRs, it was possible to differentiate the four groups. BRCA1, BRCA2 and SBC were associated with clusters of hyper-expressed miRs, and BRCAX with hypo-expressed miRs. Hsa-miR-4417 and hsa-miR-423-3p expressions (included among the eight validated miRs) differentiated 70.1 % of hereditary and non-hereditary BCs. We found miR profiles associated with tumor features like node involvement, histological grade, ER, PR and HER2 expression. Regarding molecular parameters, we only found a weak association of miRs in BC harboring losses in AURKA. We conclude that array miR expression profiles can differentiate the four study groups using FFPE BC. However, miRs expression estimated by qRT-PCR differentiates only hereditary and non-inherited BCs. The miR expression array is a simple and rapid approach that could be useful to facilitate the identification of those SBC carrying genetic or epigenetic changes in BRCA genes responsible of BRCA-like phenotype. These patients could benefit from the treatment with PARP inhibitors.

Plasma proteomics can discriminate isolated early from dual responses in asthmatic individuals undergoing an allergen inhalation challenge

PURPOSE

This proteomics study was designed to determine the utility of iTRAQ MALDI-TOF/TOF technology to compare plasma samples from carefully phenotyped mild, atopic asthma subjects undergoing allergen inhalation challenge.

EXPERIMENTAL DESIGN

Eight adult subjects with mild, allergic asthma (four early responders (ERs) and four dual responders (DRs)) participated in the allergen inhalation challenge. Blood samples were collected prior to and 2 h after the inhalation challenge. Sixteen plasma samples (two per subject), technical replicates, and pooled controls were analyzed using iTRAQ. Technical validation was performed using LC-MRM/MS. Moderated robust regression was used to determine differentially expressed proteins.

RESULTS

Although this study did not show significant differences between pre- and post-challenge samples, discriminant analysis indicated that certain proteins responded differentially to allergen challenge with respect to responder type. At pre-challenge, fibronectin was significantly elevated in DRs compared to ERs and remained significant in the multiple reaction monitoring validation.

CONCLUSIONS AND CLINICAL RELEVANCE

This proof of principle demonstration has shown that iTRAQ can uncover differences in the human plasma proteome between two endotypes of asthma and merits further application of iTRAQ to larger cohorts of asthma and other respiratory diseases.

A novel approach for biomarker selection and the integration of repeated measures experiments from two assays

BACKGROUND

High throughput 'omics' experiments are usually designed to compare changes observed between different conditions (or interventions) and to identify biomarkers capable of characterizing each condition. We consider the complex structure of repeated measurements from different assays where different conditions are applied on the same subjects.

RESULTS

We propose a two-step analysis combining a multilevel approach and a multivariate approach to reveal separately the effects of conditions within subjects from the biological variation between subjects. The approach is extended to two-factor designs and to the integration of two matched data sets. It allows internal variable selection to highlight genes able to discriminate the net condition effect within subjects. A simulation study was performed to demonstrate the good performance of the multilevel multivariate approach compared to a classical multivariate method. The multilevel multivariate approach outperformed the classical multivariate approach with respect to the classification error rate and the selection of relevant genes. The approach was applied to an HIV-vaccine trial evaluating the response with gene expression and cytokine secretion. The discriminant multilevel analysis selected a relevant subset of genes while the integrative multilevel analysis highlighted clusters of genes and cytokines that were highly correlated across the samples.

CONCLUSIONS

Our combined multilevel multivariate approach may help in finding signatures of vaccine effect and allows for a better understanding of immunological mechanisms activated by the intervention. The integrative analysis revealed clusters of genes, that were associated with cytokine secretion. These clusters can be seen as gene signatures to predict future cytokine response. The approach is implemented in the R package mixOmics (http://cran.r-project.org/) with associated tutorials to perform the analysis(a).

Regularized Partial Least Squares with an Application to NMR Spectroscopy

High-dimensional data common in genomics, proteomics, and chemometrics often contains complicated correlation structures. Recently, partial least squares (PLS) and Sparse PLS methods have gained attention in these areas as dimension reduction techniques in the context of supervised data analysis. We introduce a framework for Regularized PLS by solving a relaxation of the SIMPLS optimization problem with penalties on the PLS loadings vectors. Our approach enjoys many advantages including flexibility, general penalties, easy interpretation of results, and fast computation in high-dimensional settings. We also outline extensions of our methods leading to novel methods for non-negative PLS and generalized PLS, an adoption of PLS for structured data. We demonstrate the utility of our methods through simulations and a case study on proton Nuclear Magnetic Resonance (NMR) spectroscopy data.

A gene-phenotype network based on genetic variability for drought responses reveals key physiological processes in controlled and natural environments

Identifying the connections between molecular and physiological processes underlying the diversity of drought stress responses in plants is key for basic and applied science. Drought stress response involves a large number of molecular pathways and subsequent physiological processes. Therefore, it constitutes an archetypical systems biology model. We first inferred a gene-phenotype network exploiting differences in drought responses of eight sunflower (Helianthus annuus) genotypes to two drought stress scenarios. Large transcriptomic data were obtained with the sunflower Affymetrix microarray, comprising 32423 probesets, and were associated to nine morpho-physiological traits (integrated transpired water, leaf transpiration rate, osmotic potential, relative water content, leaf mass per area, carbon isotope discrimination, plant height, number of leaves and collar diameter) using sPLS regression. Overall, we could associate the expression patterns of 1263 probesets to six phenotypic traits and identify if correlations were due to treatment, genotype and/or their interaction. We also identified genes whose expression is affected at moderate and/or intense drought stress together with genes whose expression variation could explain phenotypic and drought tolerance variability among our genetic material. We then used the network model to study phenotypic changes in less tractable agronomical conditions, i.e. sunflower hybrids subjected to different watering regimes in field trials. Mapping this new dataset in the gene-phenotype network allowed us to identify genes whose expression was robustly affected by water deprivation in both controlled and field conditions. The enrichment in genes correlated to relative water content and osmotic potential provides evidence of the importance of these traits in agronomical conditions.

Refining developmental coordination disorder subtyping with multivariate statistical methods

BACKGROUND

With a large number of potentially relevant clinical indicators penalization and ensemble learning methods are thought to provide better predictive performance than usual linear predictors. However, little is known about how they perform in clinical studies where few cases are available. We used Random Forests and Partial Least Squares Discriminant Analysis to select the most salient impairments in Developmental Coordination Disorder (DCD) and assess patients similarity.

METHODS

We considered a wide-range testing battery for various neuropsychological and visuo-motor impairments which aimed at characterizing subtypes of DCD in a sample of 63 children. Classifiers were optimized on a training sample, and they were used subsequently to rank the 49 items according to a permuted measure of variable importance. In addition, subtyping consistency was assessed with cluster analysis on the training sample. Clustering fitness and predictive accuracy were evaluated on the validation sample.

RESULTS

Both classifiers yielded a relevant subset of items impairments that altogether accounted for a sharp discrimination between three DCD subtypes: ideomotor, visual-spatial and constructional, and mixt dyspraxia. The main impairments that were found to characterize the three subtypes were: digital perception, imitations of gestures, digital praxia, lego blocks, visual spatial structuration, visual motor integration, coordination between upper and lower limbs. Classification accuracy was above 90% for all classifiers, and clustering fitness was found to be satisfactory.

CONCLUSIONS

Random Forests and Partial Least Squares Discriminant Analysis are useful tools to extract salient features from a large pool of correlated binary predictors, but also provide a way to assess individuals proximities in a reduced factor space. Less than 15 neuro-visual, neuro-psychomotor and neuro-psychological tests might be required to provide a sensitive and specific diagnostic of DCD on this particular sample, and isolated markers might be used to refine our understanding of DCD in future studies.

Characteristic gene selection via weighting principal components by singular values

Conventional gene selection methods based on principal component analysis (PCA) use only the first principal component (PC) of PCA or sparse PCA to select characteristic genes. These methods indeed assume that the first PC plays a dominant role in gene selection. However, in a number of cases this assumption is not satisfied, so the conventional PCA-based methods usually provide poor selection results. In order to improve the performance of the PCA-based gene selection method, we put forward the gene selection method via weighting PCs by singular values (WPCS). Because different PCs have different importance, the singular values are exploited as the weights to represent the influence on gene selection of different PCs. The ROC curves and AUC statistics on artificial data show that our method outperforms the state-of-the-art methods. Moreover, experimental results on real gene expression data sets show that our method can extract more characteristic genes in response to abiotic stresses than conventional gene selection methods.