Computer-assisted Prediction On Bioactivity Of Thromboxane-a Synthase Inhibitors

Thromboxane A₂（TXA₂）is an effective inducer of platelet activation,which can cause vasoconstriction and platelet aggregation.Thromboxane-A synthase（TXS）can generate TXA₂ from prostaglandin H₂（PGH₂）.TXS is involved in the regulation of various physiological states,and inhibition of TXS activity is helpful for the treatment of cardiovascular diseases and cancer.Therefore,it is of great significance to develop TXS inhibitors.This paper aims to build a more comprehensive database of human TXS inhibitors,predict the biological activity of TXS inhibitors based on computer assistance,and explore the relationship between the structure and activity of TXS inhibitors.The main work of this paper is as follows:（1）Qualitative classification of TXS inhibitors.We established a structurally diverse database containing 526 TXS inhibitors.Self-organizing map（SOM）method and random method were used to divide the dataset into the training set and test set in a ratio of 3:1.MACCS fingerprints,ECFP4fingerprints and MOE molecular descriptors were used to characterize the molecular structure or physicochemical information of compounds.Twenty-four classification models were established based on four algorithms of support vector machine（SVM）,random forest（RF）,extreme gradient boosting（XGBoost）and deep neural network（DNN）.Model＿4A,established by DNN algorithm and MACCS fingerprints,has the best classification effect.Its prediction accuracy in the test set is 0.977,and Matthew correlation coefficient is 0.952.The distance between compound and model(d _STD-PRO)was used to characterize the application domain of the model.In the test set of Model＿4A,95.4%compounds were lower than the corresponding training set threshold(threshold_0.90=0.1440),and the prediction accuracy of these compounds were0.991.（2）Structure-activity relationship analysis of TXS inhibitors.For MACCS and ECFP4 fingerprints,we found that aromatic nitrogenous heterocyclic groups were beneficial to improve the bioactivity of TXS inhibitors.To explain this phenomenon,Alpha Fold 2 was used to construct the protein model of TXS,and the interaction between high-activity inhibitors and TXS was revealed through molecular docking.We obtained the structural characteristics of TXS inhibitors with high activity:the front part of the molecule is aromatic nitrogenous heterocyclic group,on which the nitrogen atom is vertically oriented to the heme plane,formingπ-cation interaction with Fe²⁺in the center of the heme.The middle part of the molecule is aromatic ring structure,which can form hydrophobic interaction with amino acid residues.The tail of the molecule is a carboxylic acid and other electronegative groups,which can form hydrogen bonds with amino acid residues.These structures facilitate the compound to bind to TXS and can improve the inhibitor activity.We analyzed the molecular descriptors with high correlation,and found that the complexring,and descriptors that characterize molecular polarity,lipid solubility,hydrogen bond acceptor and donor were highly correlated with the activity,which had a significant impact on the biological activity of compounds（3）Quantitative structure-activity relationship（QSAR）study of TXS inhibitors.The structure and activity values(IC₅₀ values)of 248 TXS inhibitors whose IC₅₀ were definite and measured by radiometric thin-layer chromatography assay（TLC）were collected.SOM method and random method were used to divide the quantitative dataset into training set and test set in a ratio of 3:1.2D molecular descriptors and global molecular descriptors were used to characterize the molecular structure or physicochemical information of compounds.Sixteen QSAR models were established based on SVM and RF.Model＿1D has the best prediction effect.Its coefficient of determination（R²）in the test set is 0.824,root mean squared error（RMSE）is 0.483,and mean absolute error（MAE）is 0.349.We used Williams diagram to visualize the application domain of the model.The leverage warning value of Model＿1D is0.3065,and the coverage of the model to the test set reaches 96.8%.In this paper,several TXS inhibitor classification models and TXS inhibitor regression models were constructed.These models can be used to predict TXS inhibitor activity,efficiently screen high-activity TXS inhibitors,accelerate the process of drug development,and reduce the cost of capital.The concluded structure-activity relationship can provide guidance for researchers to design high-activity TXS candidate drugs.