Research on the Compatible Features of Prescriptions in Treating COVID-19 Based on Complex Network and PCA Dimension Reduction and K-Means Algorithm Clustering

OBJECTIVE To explore the compatible characteristics of prescriptions in treating COVID-19. METHODS The traditional Chinese Medicine (TCM) treatment programs for COVID-19 issued by the National and Provincial Health Commission and National Administration of TCM， as well as the publicly published literature of Chinese and English database documents were collected. The compatible characteristics and core herbs were analyzed by data-mining methods including association rules， complex herb network， as well as drug clustering by PCA dimension reduction and K-Means algorithm. RESULTS A total of 639 prescriptions were included， involving 237 Chinese herbs. The results of frequent itemsets showed that Chinese herbs with high support were Glycyrrhizae radix et rhizoma， Armeniacae semen amarum， Ephedrae herba， Gypsum fibrosum， Scutellariae radix， Pinelliae rhizoma， Pogostemonis herba， Poria， Citri reticulatae pericarpium， Atractylodis rhizoma， Magnoliae officinalis cortex， Forsythiae fructucs， etc. The functions were dispersing and lowering lung qi， resolving dampness with aroma， as well as clearing lung and detoxifying， etc. The herb pairs and groups with high support mainly concentrated on the compatibility among Ephedrae herba， Armeniacae semen amarum， Gypsum fibrosum and Scutellariae radix， the compatibility among Poria， Pinelliae rhizoma， Pogostemonis herba and Citri reticulatae pericarpium， etc. The frequent and relevant results showed that some Chinese herbs worked closely with each other even with low support such as herb pairs composed from Aconiti lateralis radix praeparata， Corni fructus， Ginseng radix et rhizoma， Tsaoko fructus， Arecae semen， etc. The complex network results showed that the core herbs were Glycyrrhizae radix et rhizoma， Armeniacae semen amarum， Ephedrae herba， Gypsum fibrosum and Scutellariae radix， which formed complex network relationship with Magnoliae officinalis cortex， Atractylodis rhizoma， Tsaoko fructus， Arecae semen， Poria， Forsythiae fructucs， Pinelliae rhizoma， Pogostemonis herba， Citri reticulatae pericarpium， etc. 6 kinds of drug groups were gathered out with cluster analysis， which had the compatibility characteristics of clearing heat and removing toxicity in the lung， resolving dampness with aromatics， etc. CONCLSION Combined with different data mining methods， the characteristics of medication and compatibility in treating COVID-19 can be reflected from multiple angles， providing a basis for TCM in treating epidemic diseases.