Predictive modelling of colossal ATR-FTIR spectral data using PLS-DA: empirical differences between PLS1-DA and PLS2-DA algorithms

Abstract

In response to our review paper [L. C. Lee et al., Analyst, 2018, 143, 3526–3539], we present a study that compares empirical differences between PLS1-DA and PLS2-DA algorithms in modelling a colossal ATR-FTIR spectral dataset.

Rapid Prediction of Fig Phenolic Acids and Flavonoids Using Mid-Infrared Spectroscopy Combined With Partial Least Square Regression

“…One generally agreed, the chemometrics of multivariate adjustments may be used efficiently to identify the relationships between the actual concentration of targeted compounds, as determined by conventional methods, such as HPLC-DAD and predicted amounts using the mid infrared spectroscopy (FTIR spectroscopy in this case) ( Tejamukti et al, 2020 ). Partial least squares regression (PLSR) is a versatile discriminant method, that has been well documented ( Brereton and Lloyd, 2014 ; Mehmood and Ahmed, 2016 ; Lee and Jemain, 2019 ). Mathematically, the PLS model has a similar approach to PCA to project high dimensional data into a series of linear subspaces of the explanatory variables ( Lee and Jemain, 2019 ).…”

“…Partial least squares regression (PLSR) is a versatile discriminant method, that has been well documented ( Brereton and Lloyd, 2014 ; Mehmood and Ahmed, 2016 ; Lee and Jemain, 2019 ). Mathematically, the PLS model has a similar approach to PCA to project high dimensional data into a series of linear subspaces of the explanatory variables ( Lee and Jemain, 2019 ). However, the PLS model assumes a supervised learning process instead of an unsupervised learning approach in the PCA model ( Yang and Yang, 2003 ).…”

Metabolomic analysis to discriminate drug-induced liver injury (DILI) phenotypes

“…Subjectwise double cross validation (2CV) (Smit et al 2007 ) was selected for the initial assessment of three one-vs-rest models using the complete set of features. The one-vs-rest strategy decompose the original three-classes data set into three binary sub-data sets in which one class is compared with the rest of the classes included in the analysis (Lee and Jemain 2019 ) (i.e., here, cholestatic vs. the group of hepatocellular and recovered patients, hepatocellular vs. the group of cholestatic and recovered patients, and recovered vs. the group of hepatocellular or cholestatic DILI patients). Figure 3 (top) depicts the distribution of PLS predicted y values by 2CV for the assignment of the class membership in the three one-vs-rest models.…”

“…Figure 3 (top) depicts the distribution of PLS predicted y values by 2CV for the assignment of the class membership in the three one-vs-rest models. The discrimination among classes was assessed using the areas under the receiver operating characteristic (AUROCs) as figure of merit, and their statistical significance was estimated by permutation testing (100 permutations, p values < 0.05) (Lee and Jemain 2019 ). Results obtained from this analysis support the existence of relevant metabolic differences among the cholestatic, hepatocellular clinical DILI phenotypes and that of recovered patients.…”

Comparison Of Stratified And Random Iterative Sampling In Evaluation Of Pls-Da Model

“…The primary spectral dataset consisting of 1361 samples and 5401 variables has been studied and reported elsewhere (Lee, Liong, & Jemain, 2018b, 2018c, 2019a, 2019b. The practical purpose of classification model is to predict brand of unknown pen inks using based on ATR-FTIR spectrum of the ink entry.…”