Assessing feature relevance in NPLS models by VIP

Please cite this article as: A. Folch-Fortuny, J.M. Prats-Montalbán, S. Cubero, J. Blasco, A. Ferrer, VIS/NIR hyperspectral imaging and N-way PLS-DA models for detection of decay lesions in citrus fruits, Chemometrics and Intelligent Laboratory Systems (2016), AbstractIn this work an N-way partial least squares regression discriminant analysis (NPLS-DA) methodology is developed to detect symptoms of disease caused by Penicillium digitatum in citrus fruits (green mould) using visible/near infrared (VIS/NIR) hyperspectral images.To build the discriminant model a set of oranges and mandarins was infected by the fungus and another set was infiltrated just with water for control purposes. A double crossvalidation strategy is used to validate the discriminant models. Finally, permutation testing is used to select a few bands offering the best correct classification rates in the validation set. The discriminant models developed here can be potentially implemented in a fruit packinghouse to detect infected citrus fruits at their arrival from the field with affordable multispectral (3-5 channels) cameras installed in the packinglines.

show abstract

“…[44,45] 2.5. The remaining validation samples were projected onto the real model to obtain the correct classification rates.…”

Section: Accepted Manuscriptmentioning

confidence: 99%

VIS/NIR hyperspectral imaging and N-way PLS-DA models for detection of decay lesions in citrus fruits

Folch-Fortuny

Prats-Montalbán

Cubero

et al. 2016

Chemometrics and Intelligent Laboratory Systems

View full text Add to dashboard Cite

show abstract

“…Second, for more of a focus on the importance of the X-variables on the latent projection, the variable influence on projection (VIP) scores can be calculated using Eq. 6 (Favilla et al 2013). VIP is the weighted,, w 2 i combination of the sum of squares of Y explained by each latent variable, SSY i , normalised to the cumulative sum of square, SSY cum , where M is the total number of metabolites, and K is the total number of latent variables.…”

Section: Pls-da Variable Contributionmentioning

confidence: 99%

Migrating from partial least squares discriminant analysis to artificial neural networks: a comparison of functionally equivalent visualisation and feature contribution tools using jupyter notebooks

2020

View full text Add to dashboard Cite

Introduction Metabolomics data is commonly modelled multivariately using partial least squares discriminant analysis (PLS-DA). Its success is primarily due to ease of interpretation, through projection to latent structures, and transparent assessment of feature importance using regression coefficients and Variable Importance in Projection scores. In recent years several non-linear machine learning (ML) methods have grown in popularity but with limited uptake essentially due to convoluted optimisation and interpretation. Artificial neural networks (ANNs) are a non-linear projection-based ML method that share a structural equivalence with PLS, and as such should be amenable to equivalent optimisation and interpretation methods. Objectives We hypothesise that standardised optimisation, visualisation, evaluation and statistical inference techniques commonly used by metabolomics researchers for PLS-DA can be migrated to a non-linear, single hidden layer, ANN. Methods We compared a standardised optimisation, visualisation, evaluation and statistical inference techniques workflow for PLS with the proposed ANN workflow. Both workflows were implemented in the Python programming language. All code and results have been made publicly available as Jupyter notebooks on GitHub. Results The migration of the PLS workflow to a non-linear, single hidden layer, ANN was successful. There was a similarity in significant metabolites determined using PLS model coefficients and ANN Connection Weight Approach. Conclusion We have shown that it is possible to migrate the standardised PLS-DA workflow to simple non-linear ANNs. This result opens the door for more widespread use and to the investigation of transparent interpretation of more complex ANN architectures.

show abstract

“…There are a number of such variable selection methods [29], and a comparison of these is not in the scope of this paper. We have chosen to use the well-established method Variable Importance in Prediction (VIP, [30], [31]), and for SO-NPLS we use the extension of VIP for multiway regression [32].…”

Section: Identifying Important Variablesmentioning

confidence: 99%

Multi-way methods for understanding longitudinal intervention effects on bacterial communities

Maage

Steppeler

Berget

et al. 2018

Preprint

View full text Add to dashboard Cite

BackgroundThis paper presents a strategy for statistical analysis and interpretation of longitudinal intervention effects on bacterial communities. Data from such experiments often suffers from small sample size, high degree of irrelevant variation, and missing data points. Our strategy is a combination of multiway decomposition methods, multivariate ANOVA, multi-block regression, hierarchical clustering and phylogenetic network graphs. The aim is to provide answers to relevant research questions, which are both statistically valid and easy to interpret. ResultsThe strategy is illustrated by analysing an intervention design where two mice groups were subjected to a treatment that caused inflammation in the intestines. Total microbiota in fecal samples was analysed at five time points, and the clinical end point was the load of colon cancer lesions. By using different combinations of the aforementioned methods, we were able to show that: The treatment had a significant effect on the microbiota, and we have identified clusters of bacteria groups with different time trajectories.  Individual differences in the initial microbiota had a large effect on the load of tumors, but not on the formation of early-stage lesions (flat ACFs).  The treatment resulted in an increase in Bacteroidaceae, Prevotellaceae and Paraprevotellaceae, and this increase could be associated with the formation of cancer lesions. ConclusionThe results show that by applying several data analytical methods in combination, we are able to view the system from different angles and thereby answer different research questions. We believe that multiway methods and multivariate ANOVA should be used more frequently in the bioinformatics fields, due to their ability to extract meaningful components from data sets with many collinear variables, few samples and a high degree of noise or irrelevant variation.

show abstract

Assessing feature relevance in NPLS models by VIP

Cited by 91 publications

References 26 publications

VIS/NIR hyperspectral imaging and N-way PLS-DA models for detection of decay lesions in citrus fruits

VIS/NIR hyperspectral imaging and N-way PLS-DA models for detection of decay lesions in citrus fruits

Migrating from partial least squares discriminant analysis to artificial neural networks: a comparison of functionally equivalent visualisation and feature contribution tools using jupyter notebooks

Multi-way methods for understanding longitudinal intervention effects on bacterial communities

Contact Info

Product

Resources

About