The Aβ Aggregation Inhibitor And BBB Permeability Predicting Models Both Based On SVM

Alzheimer’s disease (AD) is a global problem. With the continuous development of social, material and economic conditions improved and the aging population being more and more, the number of the AD patients increased rapidly in the world, which brought a very heavy burden to the global economic, social as well as some families. Therefore R&D of drugs to treat AD has become very urgent.β Amyloid protein (Aβ) metabolic processes plays a very important role in the onset of AD. The AD treatment is expected to be fundamentally solved, if the Aβ aggregation can be effectively controlled. First of all, some support vector machine models were established to screen inhibitors of Aβ aggregation. And then from the traditional Chinese medicines Database (TCMD), some of traditional Chinese medicine were screened using the above models. Finally the permeability of the screened traditional Chinese medicine was tested by the blood-brain barrier permeability (BBB) predicting model. The final selected traditional Chinese medicine will be used for future experimental verification, and all the above work will provide some reference value for the future drug development.This study established a total of four SVM models, including two Aβ40small molecule inhibitor models, a Aβ42small molecule inhibitors model and one BBB model. The two Aβ40models are called to the Aβ40aggregation inhibitor predicting model-R and Aβ40aggregation inhibitor predicting model-T respectively, for the Aβ40Aβ40aggregation inhibitors data comes from two different experiments-Ap self seeding radio-assay and Aβ self seeding thioflavin T assay.The number of samples used to establish the BBB permeability predicting model is415. The training set accuracy of the model was92.5%, and the test set accuracy of the model was79.4%, and the average accuracy of the five-fold cross validation was80.3%.The number of samples used to build the Aβ40aggregation inhibitor predicting model-R was64. The training set accuracy of the Aβ40aggregation inhibitor predicting model-R was88.63%, and the test set accuracy of the model was75.0%, and the average accuracy of the leave-one-validation was75.0%. A total of41compounds were screened from the23033compounds of TCMD using the Aβ40aggregation inhibitor predicting model-R, then the number of traditional Chinese medicine which contain three or more kinds of compounds from the above41compounds were counted. The results were as follows:Tripterygium wilfordii, Zingiber officinale, Alpinia officinarum, and Alpinia oxyphylla. Finally only the Piper chaba (fruit), Alpinia officinarum, and Alpinia oxyphylla were in line with the requirements of the model after screened by the above BBB permeability predicting model.The number of samples used to build the Aβ40aggregation inhibitor predicting model-T was82. The training set accuracy of the Aβ40aggregation inhibitor predicting model-T was75.0%, and the test set accuracy of the model was57.7%, and the average accuracy of the leave-one-validation was71.4%. A total of41compounds were screened from TCMD using the Aβ40aggregation inhibitor predicting model-T, then the number of traditional Chinese medicine which contain three or more kinds of compounds from the above100compounds were counted. The results were as follows:Panax quinquefolium, Moschus moschiferus, Senna spectabilis (flower), Alpinia officinarum, Petiveria alliacea (including the root, stems and leaves), Alpinia oxyphylla and Cinchona ledgeriana. Finally all of the seven kinds of traditional Chinese medicine meet with the requirements of the BBB permeability predicting model after screened by the above BBB permeability predicting model.The number of samples used to build the Aβ42aggregation inhibitor predicting model was28. The training set accuracy of the Aβ42aggregation inhibitor predicting model was95.0%, and the test set accuracy of the model was62.5%, and the average accuracy of the leave-one-validation was55.00%.Only one compound from Brassica campestris, which was screened from TCMD using the Aβ42aggregation inhibitor predicting model could meet with the requirements of the BBB model after screened by the above BBB model.