Representative feature selection of molecular descriptors in QSAR modeling

Li, Jingshan; Luo, Dehan; Wen, Tengteng; Liu, Qi; Mo, Zhuofeng

doi:10.1016/j.molstruc.2021.131249

Cited by 17 publications

(16 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The PCA method implies a linear relationship between variables that will reduce the prediction accuracy of the QSAR model (P. Li et al, 2022) and PCA only retains the main components from an informatics perspective, but the main chemical features related to QSAR is ignored causing overfitting and low accuracy (J. Li et al, 2021).…”

Section: Resultsmentioning

confidence: 99%

Dimensional Reduction of QSAR Features Using a Machine Learning Approach on the SARS-Cov-2 Inhibitor Database

Azizah

Yanuar

Firdayani

2022

jppipa, pendidikan ipa, fisika, biologi, kimia

View full text Add to dashboard Cite

Quantitative Structure-Activity Relationship (QSAR) is a method that relates the chemical composition of a molecule to its biochemical, pharmaceutical and biological activities. The characteristics of a molecule's chemical constituents, such as chemical descriptors and fingerprints, are necessary to create a good QSAR model. Dimensionality reduction can alleviate the issue of several unnecessary and redundant chemical descriptors and chemical fingerprints in a high-dimensional feature-number data set by shrinking the high-dimensional original space to a low-dimensional intrinsic space. There are two categories of dimensional reduction techniques: feature extraction and feature selection. The dimension reduction approach can be utilized as a starting step in running a QSAR Virtual Screening Model on a dataset of SARS-CoV-2 inhibitor medications to create novel treatments for Covid-19 cases based on machine learning (ML) and the idea of medicinal repurposing. Fe extraction and feature selection are crucial to determining which feature sets should be applied to a specific classification process in QSAR modeling to produce reliable virtual screening results. The SARS-Cov-2 inhibitor drug database's chemical descriptor and chemical fingerprint were extracted using a simple, quick, and accurate method in this work. The total number of selected features is 12122 features. PCA, Missing values, and Random Forest are the techniques employed. The Xgboost Tree Ensemble, Naive Bayes, Support Vector Machine, Random Forest, and Deep Learning (Artificial Neural Network/Multilayer Perceptron) were used to classify the QSAR modeling on the training and test data. The Random Forest approach, when applied to all chemical descriptors and chemical fingerprint features, along with the XGBoost algorithm, yields the best feature selection results (accuracy value of 0.845 and AUC of 0.904). There are 233 characteristics for the regression QSAR approach and 273 features for the feature selection-based QSAR method of classification. Next, virtual screening of QSAR modeling of prospective drugs for Covid-19 therapy can be done utilizing the outcomes of the characteristics that have been chosen using the Random Forest approach

show abstract

Section: Resultsmentioning

confidence: 99%

Dimensional Reduction of QSAR Features Using a Machine Learning Approach on the SARS-Cov-2 Inhibitor Database

Azizah

Yanuar

Firdayani

2022

jppipa, pendidikan ipa, fisika, biologi, kimia

View full text Add to dashboard Cite

show abstract

“…The data in this paper include one-dimensional linear structure expressions of 1974 compounds in MD file and 729 MD. Generally speaking, MD cleaning is manual cleaning based on chemical intuition or characteristics observed in experiments [1]. However, manual cleaning relies heavily on professional knowledge and experience, which brings a high threshold for manual cleaning, and it is difficult to ensure the effectiveness and reliability of the results after manual cleaning and filtration, resulting in the distortion of the final effective MD samples.…”

Section: Data Preprocessing Of Molecular Descriptor Variablesmentioning

confidence: 99%

“…In this paper, LabVIEW software and manual are combined to perform cleaning. After cleaning, combined with Friedman rank analysis of variance [1], the number of 729 MD is reduced to 253.…”

Section: Data Preprocessing Of Molecular Descriptor Variablesmentioning

confidence: 99%

Optimal modeling of anti breast cancer drug candidates

Zhang¹,

Yang²

2023

HSET

View full text Add to dashboard Cite

As breast cancer is one of the most common cancers with high mortality rate in the world, this paper studies the optimal screening of anti-breast cancer candidate drugs. First, the data of 729 compound molecular descriptors are preprocessed, rough cleaned and filtered to 253, and then the cluster feature tree and correlation analysis are used to further reduce the dimension of data redundancy information, and 29 representative molecular descriptors are screened. In order to determine the importance and significance variables affecting the activity of compounds, the preliminary results were obtained by Pearson correlation analysis, and the results were further verified by Spearman correlation analysis. The stability of molecular descriptors was determined by coefficient of variation analysis, and the most representative 20 variables were selected from 29 molecular descriptors by principal component analysis. Projection pursuit model (PP) can reduce the difficulty of storage and calculation of high-dimensional data. 253 variable information reflecting biological activity are extracted with this model, and then the projection direction is optimized by using accelerated genetic algorithm (RAGA). Finally, 20 projection function values in the optimal projection direction are calculated to complete the prediction of compound activity index.

show abstract

“…Extraction of the structural activity relationship (SAR) [ 1 , 2 ] information from chemical datasets relies on the pairwise structural comparison of all toxicophore features and small molecules, which highlights the degree of the structural relationship between the compounds [ 3 , 4 , 5 , 6 ]. The Quantitative Structure–Activity Relationship (QSAR) can correlate the compound’s chemical and structural features with its physicochemical or biological properties.…”

Section: Introductionmentioning

confidence: 99%

Convolutional Neural Network Model Based on 2D Fingerprint for Bioactivity Prediction

Hamza

Bengherbia

Saeed

et al. 2022

IJMS

View full text Add to dashboard Cite

Determining and modeling the possible behaviour and actions of molecules requires investigating the basic structural features and physicochemical properties that determine their behaviour during chemical, physical, biological, and environmental processes. Computational approaches such as machine learning methods are alternatives to predicting the physiochemical properties of molecules based on their structures. However, the limited accuracy and high error rates of such predictions restrict their use. In this paper, a novel technique based on a deep learning convolutional neural network (CNN) for the prediction of chemical compounds’ bioactivity is proposed and developed. The molecules are represented in the new matrix format Mol2mat, a molecular matrix representation adapted from the well-known 2D-fingerprint descriptors. To evaluate the performance of the proposed methods, a series of experiments were conducted using two standard datasets, namely the MDL Drug Data Report (MDDR) and Sutherland, datasets comprising 10 homogeneous and 14 heterogeneous activity classes. After analysing the eight fingerprints, all the probable combinations were investigated using the five best descriptors. The results showed that a combination of three fingerprints, ECFP4, EPFP4, and ECFC4, along with a CNN activity prediction process, achieved the highest performance of 98% AUC when compared to the state-of-the-art ML algorithms NaiveB, LSVM, and RBFN.

show abstract

Representative feature selection of molecular descriptors in QSAR modeling

Cited by 17 publications

References 30 publications

Dimensional Reduction of QSAR Features Using a Machine Learning Approach on the SARS-Cov-2 Inhibitor Database

Dimensional Reduction of QSAR Features Using a Machine Learning Approach on the SARS-Cov-2 Inhibitor Database

Optimal modeling of anti breast cancer drug candidates

Convolutional Neural Network Model Based on 2D Fingerprint for Bioactivity Prediction

Contact Info

Product

Resources

About