A New Method Combining LDA and PLS for Dimension Reduction

Tang, Liang; Peng, Silong; Yi-ming, BI; Shan, Peng; Hu, Xiyuan

doi:10.1371/journal.pone.0096944

Cited by 37 publications

(25 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another way of analyzing multivariate changes is linear discriminant analysis (LDA), which attempts to identify a linear equation that combines the metabolomics variables, with different weightings, in order to predict whether animals belong to the sham or acoustic trauma groups. Partial least squares discriminant analysis (PLS-DA, equivalent to “projection to latent structures DA”) is a method that uses partial least squares regression in the discriminant analysis, which can be used to elucidate the separation between groups of variables through rotating the principal components obtained in PCA, and is particularly useful when there are more variables than observations and when there is correlation or “multicollinearity” amongst the variables (Tang et al, 2014). In orthogonal projection to latent structures DA (OPLS-DA), an extension of PLS-DA, the data from the continuous variables in the discriminant analysis function are separated into those that are predictive of the dependent variable (e.g., sham or acoustic trauma) and those that are not, resulting in enhanced diagnostics.…”

Section: Methodsmentioning

confidence: 99%

Brain Metabolic Changes in Rats following Acoustic Trauma

Zhu

et al. 2017

Front. Neurosci.

View full text Add to dashboard Cite

Acoustic trauma is the most common cause of hearing loss and tinnitus in humans. However, the impact of acoustic trauma on system biology is not fully understood. It has been increasingly recognized that tinnitus caused by acoustic trauma is unlikely to be generated by a single pathological source, but rather a complex network of changes involving not only the auditory system but also systems related to memory, emotion and stress. One obvious and significant gap in tinnitus research is a lack of biomarkers that reflect the consequences of this interactive “tinnitus-causing” network. In this study, we made the first attempt to analyse brain metabolic changes in rats following acoustic trauma using metabolomics, as a pilot study prior to directly linking metabolic changes to tinnitus. Metabolites in 12 different brain regions collected from either sham or acoustic trauma animals were profiled using a gas chromatography mass spectrometry (GC/MS)-based metabolomics platform. After deconvolution of mass spectra and identification of the molecules, the metabolomic data were processed using multivariate statistical analysis. Principal component analysis showed that metabolic patterns varied among different brain regions; however, brain regions with similar functions had a similar metabolite composition. Acoustic trauma did not change the metabolite clusters in these regions. When analyzed within each brain region using the orthogonal projection to latent structures discriminant analysis sub-model, 17 molecules showed distinct separation between control and acoustic trauma groups in the auditory cortex, inferior colliculus, superior colliculus, vestibular nucleus complex (VNC), and cerebellum. Further metabolic pathway impact analysis and the enrichment overview with network analysis suggested the primary involvement of amino acid metabolism, including the alanine, aspartate and glutamate metabolic pathways, the arginine and proline metabolic pathways and the purine metabolic pathway. Our results provide the first metabolomics evidence that acoustic trauma can induce changes in multiple metabolic pathways. This pilot study also suggests that the metabolomic approach has the potential to identify acoustic trauma-specific metabolic shifts in future studies where metabolic changes are correlated with the animal's tinnitus status.

show abstract

Section: Methodsmentioning

confidence: 99%

Brain Metabolic Changes in Rats following Acoustic Trauma

Zhu

et al. 2017

Front. Neurosci.

View full text Add to dashboard Cite

show abstract

“…A partial least square (PLS) regression with a backward feature selection was performed to select features that predicted CP from the large set of 990 movement features without overfitting the model [24]. The selected movement features in each 5 second period were clustered into 5 composite scores which were used in a linear discriminative analysis (LDA) to classify movements typically found in children with or without CP (Appendix C) [25]. To avoid overfitting of the CIMA model, the dataset was divided into training, validation, and test sets in a double cross-validation procedure (Appendix D).…”

Section: Cp Prediction Model and Validity Of The Modelmentioning

confidence: 99%

Machine Learning of Infant Spontaneous Movements for the Early Prediction of Cerebral Palsy: A Multi-Site Cohort Study

Ihlen

Støen

Boswell

et al. 2019

JCM

View full text Add to dashboard Cite

Background: Early identification of cerebral palsy (CP) during infancy will provide opportunities for early therapies and treatments. The aim of the present study was to present a novel machine-learning model, the Computer-based Infant Movement Assessment (CIMA) model, for clinically feasible early CP prediction based on infant video recordings. Methods: The CIMA model was designed to assess the proportion (%) of CP risk-related movements using a time-frequency decomposition of the movement trajectories of the infant's body parts. The CIMA model was developed and tested on video recordings from a cohort of 377 high-risk infants at 9-15 weeks corrected age to predict CP status and motor function (ambulatory vs. non-ambulatory) at mean 3.7 years age. The performance of the model was compared with results of the general movement 2 of 17 assessment (GMA) and neonatal imaging. Results: The CIMA model had sensitivity (92.7%) and specificity (81.6%), which was comparable to observational GMA or neonatal cerebral imaging for the prediction of CP. Infants later found to have non-ambulatory CP had significantly more CP risk-related movements (median: 92.8%, p = 0.02) compared with those with ambulatory CP (median: 72.7%). Conclusion: The CIMA model may be a clinically feasible alternative to observational GMA.

show abstract

“…Another way of analyzing multivariate changes is linear discriminant analysis (LDA), which attempts to identify a linear equation that combines the metabolomic variables, with different weightings, in order to predict if the animals belong to a certain group. Partial least squares discriminant analysis (PLS-DA, equivalent to "projection to latent structures DA") is a method that uses partial least squares regression in the discriminant analysis and can be used to elucidate the separation between groups of variables [28]. In orthogonal projection to latent structures DA (OPLS-DA), an extension of PLS-DA, the data from the continuous variables in the discriminant analysis function are separated into those that are predictive of the dependent variable (e.g., WT or db/ db) and those that are not, resulting in enhanced diagnostics.…”

Section: Sample Preparation For Metabolomic Analysismentioning

confidence: 99%

The aldose reductase inhibitor epalrestat exerts nephritic protection on diabetic nephropathy in db/db mice through metabolic modulation

Gao

Yang

et al. 2018

Acta Pharmacol Sin

View full text Add to dashboard Cite

Epalrestat is an inhibitor of aldose reductase in the polyol pathway and is used for the management of diabetic neuropathy clinically. Our pilot experiments and accumulated evidences showed that epalrestat inhibited polyol pathway and reduced sorbitol production, and suggested the potential renal protection effects of epalrestat on diabetic nephropathy (DN). To evaluate the protective effect of epalrestat, the db/db mice were used and exposed to epalrestat for 8 weeks, both the physiopathological condition and function of kidney were examined. For the first time, we showed that epalrestat markedly reduced albuminuria and alleviated the podocyte foot process fusion and interstitial fibrosis of db/db mice. Metabolomics was employed, and metabolites in the plasma, renal cortex, and urine were profiled using a gas chromatography-mass spectrometry (GC/MS)-based metabolomic platform. We observed an elevation of sorbitol and fructose, and a decrease of myo-inositol in the renal cortex of db/db mice. Epalrestat reversed the renal accumulation of the polyol pathway metabolites of sorbitol and fructose, and increased myo-inositol level. Moreover, the upregulation of aldose reductase, fibronectin, collagen III, and TGF-β1 in renal cortex of db/db mice was downregulated by epalrestat. The data suggested that epalrestat has protective effects on DN, and the inhibition of aldose reductase and the modulation of polyol pathway in nephritic cells be a potentially therapeutic strategy for DN.

show abstract

A New Method Combining LDA and PLS for Dimension Reduction

Cited by 37 publications

References 41 publications

Brain Metabolic Changes in Rats following Acoustic Trauma

Brain Metabolic Changes in Rats following Acoustic Trauma

Machine Learning of Infant Spontaneous Movements for the Early Prediction of Cerebral Palsy: A Multi-Site Cohort Study

The aldose reductase inhibitor epalrestat exerts nephritic protection on diabetic nephropathy in db/db mice through metabolic modulation

Contact Info

Product

Resources

About