Multivariate calibration maintenance and transfer through robust fused LASSO

et al. 2016

A new quantitative structure-activity relationship (QSAR) of the inhibition of mild steel corrosion in 1 M hydrochloric acid using furan derivatives was developed by proposing two-stage sparse multiple linear regression. The sparse multiple linear regression using ridge penalty and sparse multiple linear regression using elastic net (SMLRE) were used to develop the QSAR model. The results show that the SMLRE-based model possesses high predictive power compared with sparse multiple linear regression using ridge penalty-based model according to the mean-squared errors for both training and test datasets, leave-one-out internal validation (Q 2 int = 0.98), and external validation (Q 2 ext = 0.95). In addition, the results of applicability domain assessment using the leverage approach reveal a reliable and robust SMLRE-based model. In conclusion, the developed QSAR model using SMLRE can be efficiently used in the studies of corrosion inhibition efficiency.

Section: Methodsmentioning

confidence: 99%

“…To handle the high dimensionality problem, selection of the relevant descriptors is an essential step in constructing QSAR models. Sparse regression methods are an attractive framework that have been adapted and gained popularity for performing descriptor selection and QSAR model estimation in high‐dimensional data simultaneously .…”

Section: Introductionmentioning

confidence: 99%

Quantitative structure–activity relationship model for prediction study of corrosion inhibition efficiency using two‐stage sparse multiple linear regression

Al-Fakih

et al. 2016

“…It has become an increasingly essential research area, especially with the computational statistical methods. In chemometrics, a well‐known molecular descriptor selection method is the PLR, which is simultaneously performing descriptor selection and QSAR classification . This method is of more interest for researchers because it continuously shrinking the descriptor coefficients toward 0 and it is computationally feasible.…”

Section: Methodsmentioning

confidence: 99%

“…Penalized methods offer several advantages over the classical subset selection methods alternative, including stability, interpretability, and the ability to handle high‐dimensional data. Penalized methods have been proposed in several works and establish one of the most widely used methods for molecular descriptor selection because of their satisfactory empirical and theoretical properties.…”

Section: Introductionmentioning

confidence: 99%

A novel molecular descriptor selection method inQSARclassification model based on weighted penalized logistic regression

2017

Molecular descriptor selection is a pivotal tool for quantitative structure–activity relationship modeling. This paper proposes a novel molecular descriptor selection method on the basis of taking into account the information of the group type that the descriptor belongs to. This descriptor selection method is on the basis of combining penalized logistic regression with 2‐sample t test. The proposed method can perform filtering and weighting simultaneously. Specifically, 2‐sample t test is employed as filter method by removing the descriptor which is not show statistically significant difference. On the other hand, a weighted penalized logistic regression is used by assigning a weight depending on the 2‐sample t test value inside the descriptor type block. The proposed method is experimentally tested and compared with state‐of‐the‐art selection methods. The results show that our proposed method is simpler and faster with efficient classification performance.

“…Among them, are LASSO (L1‐norm) , SCAD , elastic net , adaptive LASSO , and adaptive elastic net . In QSAR studies, LASSO has been applied and compared with other methods .…”

Section: Methodsmentioning

confidence: 99%

High‐dimensional quantitative structure–activity relationship modeling of influenza neuraminidase a/PR/8/34 (H1N1) inhibitors based on a two‐stage adaptive penalized rank regression

Al-Fakih

2015

Outliers in the biological activity variable or the heavy tailed distribution of the error are often encountered in practice. Under these circumstances, the quantittative structure-activity relationship (QSAR) model using multiple linear regression is not efficient. In this paper, a two-stage adaptive penalized rank regression is proposed for constructing a robust and efficient high-dimensional QSAR model of influenza virus neuraminidase A/PR/8/34 (H1N1) inhibitors. The results demonstrate the effectiveness of our proposed method in simultaneously estimating a robust QSAR model and selecting informative molecular descriptors. Furthermore, the results prove that the proposed method can significantly encourage the grouping effect. The proposed method, because of the high predictive ability and robustness, could be a useful method in high-dimensional QSAR modeling.