A Curated Database Of Antioxidant Molecular And Its QSAR Analysis

Free radicals attack almost all types of biochemicals,such as lipids,proteins,DNA,etc.,which are free-radical chain reactions.Excessive free radicals in humans cause oxidative stress,which can lead to the development of various types of systemic diseases.The traditional antioxidant screening method requires a lot of manpower and material resources.Therefore,it is urgent to establish a statistically significant and predictive QSAR model to quickly and accurately predict the antioxidant activity of the screened compounds.The purpose of this study was to establish a high-quality in vitro antioxidant database,and based on this,a stable and reliable QSAR model was established by using the principles of chemical informatics to establish a quantitative structure-activity relationship between the ability of compounds to scavenge DPPH free radicals and corresponding structural features.The model is used to predict the antioxidant activity of the compound.Used to screen new potential antioxidants.Compounds with antioxidant activity were collected from the Pub Chem database and literature,and by removing duplicate values,inorganics and discrete values.The complete antioxidants DB database of 799 compounds was retained.The antioxidant molecules in this database have p IC50 activity values ranging from 3.2 to 5.5,log P(o/w)octanol/water partition coefficients ranging from 4.5 to 20.3,and antioxidant molecules ranging from 110 to 2809.9 g/mol.The natural source of antioxidant compounds in the database accounted for 37.3%.The molecular weight is between 200 and 600 g/mol,and the log P(o/w)contains the most antioxidant molecules between 0 and 5.Most of the synthetic antioxidants are lipophilic.Natural antioxidants have a large molecular weight.The main compound classes in antioxidants DB are benzimidazole derivatives,coumarin derivatives,pyrazole derivatives,pyridine derivatives,thiazole derivatives and the like.The physicochemical,biologically active information contained in the database,such as functional groups,antioxidant activity intervals,molecular weight,hydrogen bond donors,hydrogen bond acceptors,and molecular types,contribute to a comprehensive understanding of antioxidant molecules.Through this database,feature-based queries such as compound similarity search,structure search,etc.can be performed.The use of a database at the same time allows researchers to quickly understand the current state of research on antioxidants for further data analysis and prediction.In the development of QSAR,molecular descriptors serve as a bridge between activity and small molecule structure and are the basis of QSAR analysis.A total of 322 2D and 3D molecular descriptors were generated using MOE software.Unrelated redundancy descriptors affect prediction accuracy.There are 300 descriptors that are not related to antioxidant activity by PCA and collinearity analysis,and the important 22 descriptors are reserved as PEOE_VSA-3,vsurf_CW8,E_sol,a_don,vsurf_HB8,Slog P_VSA0,h_p Ka,and the like.Based on 22 molecular descriptors,five QSAR models were established using PLS,SWR-MLR,SVR,RF,and k NN algorithms.Through the internal 10 fold cross-validation of the training set,Q2 is greater than 0.64 and RMSEcv is less than 0.23.For the test set model,R2 is greater than 0.67,and RMSE is less than 0.22.The model has good fit and predictability.The internal 10 fold cross-validation of the RF model has the best effect Q2 equals 0.70,and the RMSEcv is 0.20 is smaller than the other models.The predictive power for the test set k NN model is preferably R2 equal to 0.74 and RMSE equal to 0.19.The developed model has predictive power and can be used to predict the antioxidant activity of a group of compounds.The correlation between the predicted and experimental values of the model is greater than 0.5.The PLS model has the best prediction effect on this set of external validation sets,R is equal to 0.63.Through the research of this project,a database was established,which laid the foundation for the research of antioxidants using the principle of chemical informatics,and further promoted the research and screening of antioxidants through the principles and methods of chemical informatics.Despite our best efforts,antioxidants DB is not an exhaustive repository of all antioxidant molecules.In the future,we intend to increase the coverage of antioxidant molecules and look for their potential impact on human health.We will continue to increase the coverage of antioxidant molecular databases,increase model prediction capabilities,and conduct studies on the potential effects of antioxidants on human health.In order to obtain more information about the physicochemical determinants of antioxidant capacity,it is worthwhile to expand the chemical space of our known antioxidant molecules.