Chemometrics and Intelligent Laboratory Systems最新文献

Unbalanced sample groups tend to yield models with a higher prevalence of predominant classes. A sample group with balanced classes contributes to the development of more robust models with improved predictive capability to classify classes equally. In the literature, two methodologies for sample balancing can be found: elimination (undersampling) and synthetic sample generation (oversampling). Undersampling methodologies result in the loss of real samples, while oversampling methods may introduce issues related to adding non-real signals to the original spectra. To overcome these challenges, this paper aimed to utilize Principal Component Analysis (PCA) for the generation of virtual samples (synthetic samples and artificial outliers) to balance data in multivariate classification models. The proposed methodology was applied to data from mid-infrared spectroscopy (MIR) and high-resolution mass spectrometry (HRMS) with Partial Least Squares Discriminant Analysis (PLS-DA) and Support Vector Machine (SVM) models. The constructed models demonstrate that the addition of virtual samples enhances performance parameters (e.g., false negative rate, false positive rate, accuracy, sensitivity, specificity, among others) compared to unbalanced models, while also mitigating overfitting (a problem found in unbalanced models). Performance parameters exhibited a more significant improvement percentage using the non-linear model (SVM) compared to the linear model (PLS-DA). Furthermore, the created virtual spectra do not introduce new signals, i.e., original, and virtual spectra exhibit a similar spectral profile, differing only in the intensity levels. Finally, all models demonstrated good predictive capability according to permutation testing for the binary model developed in this work, limiting the rate of class permutation retention (between 40 % and 60 % of the y-vector remained in the original class). All created models exhibited accuracy values higher than the accuracy distribution of models with permuted classes for the test group.

Dealing with missing data poses a challenge in Principal Component Analysis (PCA) since the most common algorithms are not designed to handle them. Several approaches have been proposed to solve the missing value problem in PCA, such as Imputation based on SVD (I-SVD), where missing entries are filled by imputation and updated in every iteration until convergence of the PCA model, and the adaptation of the Nonlinear Iterative Partial Least Squares (NIPALS) algorithm, able to work skipping the missing entries during the least-squares estimation of scores and loadings. However, some limitations have been reported for both approaches. On the one hand, convergence of the I-SVD algorithm can be very slow for data sets with a high percentage of missing data. On the other hand, the orthogonality properties among scores and loadings might be lost when using NIPALS.

To solve these issues and perform PCA of data sets with missing values without the need of imputation steps, a novel algorithm called Orthogonalized-Alternating Least Squares (O-ALS) is proposed. The O-ALS algorithm is an alternating least-squares algorithm that estimates the scores and loadings subject to the Gram-Schmidt orthogonalization constraint. The way to estimate scores and loadings is adapted to work only with the available information.

In this study, the performance of O-ALS is tested and compared with NIPALS and I-SVD in simulated data sets and in a real case study. The results show that O-ALS is an accurate and fast algorithm to analyze data with any percentage and distribution pattern of missing entries, being able to provide correct scores and loadings in cases where I-SVD and NIPALS do not perform satisfactorily.

Functional MRI is, currently, the most sensitive technique in breast cancer for detecting early tumors, and perfusion (DCE-MRI) has become the most important sequence to depict and characterize angiogenesis and neovascularization. In this work, we propose the use of new biomarkers that are related to clear physiological phenomena, obtained from MCR-ALS as an alternative to curve-based pseudo-biomarkers and pharmacokinetics models. In order to provide a discrimination and prediction model between healthy tissue and cancer, we propose using PLS-DA with double cross-validation (2CV) and variable selection, repeated several times and obtaining excellent average results for the performance indexes (f-score: 0.9149, MCC: 0.8538, AUROC: 0.8794). After selecting the optimal prediction model, a unique probabilistic map called “virtual biopsy” that shows in different colors the probability that each pixel of the image has a tumor behavior is obtained, helping the specialist with the identification and characterization of breast tumors with only one easy-to-interpret biomarker map.

Time-series prediction is of great practical value in industrial scenarios such as rotary kilns, especially for long sequence time-series prediction. Accurate long sequence NOx emission predictions help us monitor rotary kiln operations in advance to plan and control NOx emissions according to emission policies and production requirements. However, in actual industrial scenarios, the NOx emission pattern is dominated by long-term trends rather than simply repetitive patterns. Existing NOx prediction models are not effective in capturing long-term dependencies. Therefore, this paper proposes a novel model based on Transformer to solve this problem. First, we propose a novel series decomposition architecture based on LSTM and self-attention, which is embedded inside the Transformer. The architecture allows self-attention at the sub-series level and provides short-term trend and position information. In addition, the model designs a one-step inference structure to improve the error accumulation phenomenon under traditional inference methods for long sequence prediction and reduce the inference time. We conducted extensive experiments on two real-world datasets with different sampling intervals, which validated the model’s effectiveness. It achieves a relative improvement of 53.2% and 43.4% in prediction accuracy compared to popular NOx emission prediction methods.

Tyre materials are complex products, as they are prepared using a number of raw materials, each of them with its specific chemical composition and functionality in the final product. It is, therefore, of crucial importance to avoid mislabeling errors and even to verify the compliance of raw materials entering the factory.

The present study proposes a strategy that makes use of near infrared (NIR) spectroscopy combined with chemometrics for raw material identification (RMID) and compliance verification of the most common raw materials used in the tyre industry. In particular, the chemometric model developed consists of a global hierarchical classification model, which combines nested PLS-DA nodes for RMID and SIMCA nodes for compliance verification, in a two-step approach.

The global model showed satisfactory results, as a 100 % of total correct predictions and a sensitivity higher than 90 % in the test set were obtained for most of the classes of interest.

The strategy obtained has the final goal of being directly applied on the raw materials at their receiving stage in factory, with the double advantage of minimizing the risk of mislabeling and, at the same time, decreasing the number of suspicious samples that need to be analyzed in the laboratory, by means of traditional methods, for verifying their compliance.

Introduction

Assessing repeatability and reproducibility in analytical chemistry is commonly based on parametric dispersion indicators, such as relative standard deviation and standard deviation, calculated for each detected variable using repeated measurements of Quality Control (QC) samples collected throughout the data acquisition sequence. However, their reliability strongly relies on the assumption of normality distribution. Knowing that analytical variability is conditional to many sources, the use of such parametric estimators is not always suitable. There is therefore a need for robust indicators of data quality independent of central values and any parametric assumption.

Methods

Three specific indicators were developed: (i) intra-group dispersion, based on the median area of the convex hull of QC samples within an analytical batch; (ii) inter-group dispersion, defined as the gradient of the deviation between analytical batches; and (iii) dispersion index. Mathematical properties of these indicators, including positivity, stability, and translation invariance, were then evaluated using synthetic data under normal and non-normal distributions. Finally, the relevance of these indicators and the associated visualization methods were highlighted based on a metabolomics case study involving liquid chromatography coupled to mass spectrometry measurements of the NIST SRM1950 reference material analyzed over more than one year within different projects.

Results

The proposed indicators were shown to be translation invariant and always positive, while first investigations performed on synthetic data revealed a high stability for multiplication. Moreover, their application to experimental data revealed specific behaviors depending on the characteristics of the signal associated with the different detected analytes, showing their ability to capture the variability observed either in parametric or non-parametric conditions. Moreover, this investigation showed different structures of sensitivity to analytical variability all along the data processing steps. The proposed indicators also allowed a visualization of the analytical drift in two dimensions, to facilitate result interpretation.

Conclusion

These indicators open the way to a better and more robust assessment of repeatability and reproducibility but also to improvements of long-term data comparability involving suitability testing.

One of the main issues affecting the effectiveness of the quantitative structure-activity relationship (QSAR) classification techniques in chemometrics is high dimensionality. Applying feature selection is a critical procedure that determines the most relevant and important aspects of a dataset. It improves the effectiveness and accuracy of prediction models by effectively lowering the number of features. This decrease increases classification accuracy, reduces computing strain, and improves overall performance. Recently, the golden jackal optimization (GJO) algorithm was introduced, which has been successfully used to solve various continuous optimization issues. Therefore, this study proposes an improvement in the GJO algorithm employing chaotic maps, abbreviated as CGJO, to enhance the exploration and exploitation capability of the GJO algorithm in picking the essential descriptors in QSAR classification models with high classification accuracy and less computation time. Experimental findings based on four different high-dimensional chemical datasets show that the proposed CGJO algorithm can maximize classification accuracy while simultaneously decreasing the number of chosen descriptors and lowering the time required for computing. Thus, the proposed algorithm can be useful for chemical data classification in other QSAR modeling.

Recent data from the World Health Organization reveals that in 2023, 38.8 million people were living with HIV. Within this population, there were 1.5 million new cases and 650 thousand deaths attributed to the disease. This study employs an integrated approach involving QSAR-based machine learning models, molecular docking, and molecular dynamics simulations to identify potential compounds for inhibiting the bioactivity of the CC chemokine receptor type 5 (CCR5) protein, a key entry point for the HIV virus. Using non-redundant experimental data from the CHEMBL database, 40 different machine learning algorithms were trained and the top four models (XGBoost, Histogram based gradient Boosting, Light Gradient Boosted Machine, and Extra Trees Regression) were utilized to predict anti-HIV bioactivity for 37 billion compounds in the ZINC-22 database. The screening resulted in the identification of 124 new anti-HIV drug candidates, confirmed through molecular docking and dynamics simulations. The study underscores the therapeutic potential of these compounds, paving the way for further in vitro and in vivo investigations. The convergence of machine learning and experimental findings presents a promising avenue for significant advancements in pharmaceutical research, particularly in the treatment of viral diseases such as HIV. To guarantee the reproducibility of our study, we have made the Python code (google colab) and the associated database available on GitHub. You can access them through the following link: GitHub Link: https://github.com/AlexandreCOBRE/code.

Macroscopic X-ray fluorescence (MA-XRF) datasets are analyzed using Artificial Neural Networks. Specifically, Convolutional Neural Networks (CNNs) are trained by coupling the spectra acquired during the MA-XRF scan of two religious panel paintings (“icons”) with the associated Ground-Truth counts per characteristic transition line, as they are extracted by X-ray fluorescence fundamental parameters analysis. In total, twenty thousand XRF spectra were used for the CNN training. The trained neural networks were applied to analyze millions of MA-XRF spectra acquired during the scan of religious painting panels by computing the counts per pixel of X-ray characteristic transition lines and creating the elemental transition maps. Comparison of the CNN extracted results to the Ground-Truth (GT) shows remarkable agreement. The successful MA-XRF datasets analysis applying the CNN method paves an analytical path to the direction of the auto-identification of spectral lines, offering the means for the non-experienced XRF analyst to provide a state-of-the-art analysis and supporting the experienced user not to overlook hardly resolved transition lines.

The browning process of food products if often formed upon cutting and damage during their processing, transport, and storage, amongst other potential sources and reasons. Enzymic browning can be mainly due to polyphenol oxidase (PPO) and peroxidase (POD) enzymes. Visible/near-infrared (Vis/NIR) imaging spectroscopy in the range of 350–1150 nm was used in this study for automatic and non-destructive evaluation of PPO and POD activity levels in three bell pepper varieties (red, yellow, orange; N = 30), with a total of 30 inputs samples in each variety. The spectral data were then modeled by the partial least squares regression (PLSR) throughout the whole spectral range, without using any subset of the most effective wavelength (EW) values. Regression determination coefficient (R²) values for the estimation (prediction) of POD enzyme activity levels were 0.794, 0.772, and 0.726 for red, yellow, and orange bell peppers, respectively, all over the validation set. At the same time, the activity levels of PPO enzyme over bell peppers showed R² values of 0.901, 0.810, and 0.859, for red, yellow, and orange bell peppers, respectively, all over the validation set. In addition, a combination of support vector machine (SVM) with either genetic algorithms (GA), particle swarm optimization (PSO), ant colony optimization (ACO), or imperialistic competitive algorithms (ICA) hybrid machine learning (ML) techniques were used to select the optimal (discriminant) spectral EW wavelength values, and regression performance was consistently improved, to judge from higher regression fit R² values. Either 14 or 15 EWs were computed and selected in order of their discriminative power using previously mentioned ML techniques. The hybrid SVM-PSO method resulted the best one in the process of selecting the most effective wavelength values (nm). On the other hand, three regression methods comprising PLSR, multiple least regression (MLR), and neural network (NN), were employed to model the SVM-PSO selected EWs. The ratio of performance to deviation (RPD), the R² and the root mean square error (RMSE), over the test set, for the non-linear NN regression method exhibited better results as compared to the other two regression methods, being closely followed by PLSR, and therefore NN regression method was selected as the best approach for modeling the most effective spectral wavelength values in this study.