Reveal the mechanisms of prescriptions for liver cancer' treatment based on two illustrious senior TCM physicians

Abstract

OBJECTIVE:

To explore the core herbs of two illustrious senior Traditional Chinese Medicine (TCM) Physicians for the treatment of liver cancer, and to further clarify the gene targets and pathway mechanisms of liver cancer that the core prescription (CP) may regulate.

METHODS:

We used the patient information of two illustrious senior TCM physicians from the Affiliated Hospital of Shandong University of Traditional Chinese Medicine and the Affiliated Hospital of Capital Medical University as the database. The CP was analyzed using the community network algorithm. The pathway mechanism was analyzed using network pharmacology method. And the prognostic survival genes were identified using Single factor cox regression analysis. Integrative Pharmacology-based Research Platform of Traditional Chinese Medicine (TCMIP), Herb, Online Mendelian Inheritance in Man (OMIM), Genecards, Kyoto Encyclopedia of Genes and Genomes (KEGG), Genomic Data Commons (GDC), The Cancer Genome Atlas (TCGA), Gephi and R were used to mine CP, building a pathway network diagram. All the analyses were visualized.

RESULTS:

We found that the CP consistes of Huangqi (Radix Astragali Mongolici), Danshen (Radix Salviae Miltiorrhizae), Shuihonghuazi (Polygoni Orientalis Fructus), Baihuasheshecao (Herba Hedyotdis), Banzhilian (Herba Scutellariae Barbatae), and Ezhu (Rhizoma Curcumae Phaeocaulis), which were attributed to the two physicians respectively. The CP played an anti-cancer role through such pathways as signal transduction, energy metabolism, immune system and other pathways, covering a total of 112 pathways and 176 herb-disease-related genes. Fourteen genes have a significant impact on the prognosis and survival of liver cancer.

CONCLUSIONS:

Based on the liver cancer cases of two illustrious senior TCM physicians, we obtained the CP through data mining. The CP may mainly exert anti-cancer effects by inhibiting inflammatory response, angiogenesis, and enhancing the body's immune response. We screened out 14 genes in the CP that may be related to the prognosis of liver cancer, and these genes may play an important regulatory role in the prognosis of liver cancer.

1. INTRODUCTION

Liver cancer is the sixth most diagnosed cancer and the fourth leading cause of cancer death globally.¹ Hepatocellular carcinoma (HCC) is the main histological subtype of primary liver cancer, accounting for 80% of all liver cancers. In China, liver cancer is the second most lethal cancer.² The incidence and mortality of HCC have been on the rise since the 1990s.³ It is estimated that in 2030 more than 1 million people worldwide will die of HCC.⁴ HCC is associated with multiple risk factors —hepatitis virus infection, environmental toxins, oxidative stress, chronic inflammation, hepatotoxic drugs — and these triggers are involved in the development of chronic inflammation, fibrosis, and carcinogenesis.^5,6 Western medicine are the main strategies for the treatment of HCC (such as surgical resection, liver transplantation, radiofrequency, chemotherapy, and targeted molecular therapy), but the overall prognosis of HCC patients remains poor.^7,8

In TCM theory, the occurrence of tumors is related to the imbalance between Yin and Yang (immune imbalance), insufficiency of Qi and blood, and the stimulation of external harmful factors.⁹ Traditional Chinese Medicine (TCM) uses herbs, acupuncture and moxibustion, diet, exercise and massage to prevent and treat tumors by restoring the balance of the individual's internal environment. The treatment of cancer by TCM, including HCC, is based on knowledge gained from clinical experience and has achieved curative effects.¹⁰ These experience has been inherited and preserved by TCM doctors¹¹ . TCM practitioners have extensive experience in diagnosing and treating this potentially lethal tumor. Among them, the most commonly used herbs are those for tonic, heat-clearing, blood-activating and stasis-resolving.⁹ TCM has been proven to improve the clinical symptoms of HCC patients such as ascites and jaundice,^6,12 improve the quality of life and survival rate of HCC patients,^11,13,14 and reduce the adverse reactions of chemotherapy drugs.¹⁵ Recent developments in molecular and cellular biology have provided important insights into the mechanisms by which TCM intervenes in HCC, and these studies have focused on influencing HCC proliferative activity,¹⁶ apoptosis,¹⁷ metastasis,¹⁸ angiogenesis,¹⁹ immune responses,²⁰ and multidrug resistance.²¹

The latest edition of the International Classification of Diseases (ICD-11), approved by the World Health Organization (WHO), is the first authoritative body to include a chapter on TCM,²² which reflects the broad influence of TCM on a global scale. In China, TCM is the most widely used form of complementary and alternative medicine,^17,23 and can benefit HCC patients whether used alone or in combination.¹³ Although there are documented and proven efficacy of TCM in the treatment of liver cancer,^24,25 the actual anti-cancer mechanism of TCM prescription remains unclear. In recent years, systems pharmacology, especially its subfield network pharmacology, has been serving as a useful tool for further understanding of drug action.²⁶ In addition, scientists discovered more potential oncogenes and anti-cancer genes in HCC. Currently, different gene experiments for HCC have achieved valuable results in vitro or in vivo.²⁷ The reported biomarker showed better specificity than alpha-fetoprotein (AFP), which is widely used in HCC.²⁸ In addition to the existing traditional strategies, gene therapy has also shown promising prospects in the treatment of HCC, exerting a positive impact on the survival and prognosis of HCC.²⁹

Illustrious senior TCM Physicians are the outstanding representatives among TCM physicians. Their treatment experience reflects the current high level of TCM treatment for liver cancer in now. Two illustrious senior TCM physicians participating in this study have been engaged in liver cancer treatment for more than 40 years.^30,31 They have achieved positive clinical results and have been widely recognized by the community. In this study, treatment prescriptions of TCM were collected from the real word in order to represent the actual level of two illustrious senior TCM Physicians. The objective of this study is to explore the core herbs of two illustrious senior TCM Physicians for the treatment of liver cancer, and to further clarify the gene targets and pathway mechanisms of liver cancer that the core prescription (CP) may regulate. To some extent, this study can advance the medical community's understanding and acceptance of TCM therapy, and enrich the treatment strategies for liver cancer. Figure 1 shows the workflow diagram.

Figure 1. Workflow diagram.

TCMIP: Integrative Pharmacology-based Research Platform of Traditional Chinese Medicine; OMIM: Online Mendelian Inheritance in Man; KEGG: Kyoto Encyclopedia of Genes and Genomes; GDC: Genomic Data Commons; TCGA: The Cancer Genome Atlas.

2. MATERIALS AND METHODS

2.1. Clinical data collection

We retrospectively collected 101 cases of liver cancer patients treated by two illustrious senior TCM physicians from January 2009 to December 2020 in Beijing Hospital of TCM Affiliated to Capital Medical University and Affiliated Hospital of Shandong University of TCM. We collected general clinical information and Chinese medicine prescription information of these cases, including the patient's age, gender, course of disease, diagnosis, and prescription of Chinese medicine. All data was entered into Microsoft Excel 2016 to establish a database for the treatment of liver cancer by two physicians. The prescription herbs were standardized in the following manner. In accordance with the Chinese Pharmacopoeia of the People's Republic of China (2020 edition) and ‘Chinese Materia Medica’, the properties, flavors and meridian tropism of the herbs in the prescription were further supplemented. Two researchers were assigned to complete the information entry work separately. This study was reviewed by the Ethics Committee of Beijing University of TCM with the ethics approval number 2019BZHYLL0204, and was implemented in strict accordance with the ethical system.

2.2. Prescription analysis

Through the descriptive analysis of the dosages of all prescriptions, the properties, flavors, and the meridian tropism of the herbs involved, overall effect of the prescription can be summarized and analyzed. New prescriptions could be identified by applying complex network analysis methods. GraphPad Prism9, RStudio 1.4.1717 and gephi0.9.2 were used for visual display such as description analysis.

2.2.1. Descriptive analysis: dosage, herbal properties, and flavors

The descriptive analysis of characteristics of herbs includes three parts: property, flavor and meridian tropism. The four properties include five items of cold, heat, warm, cool and mild, and the five flavors includes seven items of sour, bitter, sweet, pungent, salty, astringent, and bland. Meridian tropism involves twelve meridians, which are divided into two categories: Yin meridian and Yang meridian. The top 20 core herbs were selected for visual analysis of their properties and flavors in all cases. RStudio software was used for analysis and graph drawing.

2.2.2. Complex network analysis based on gephi

Gephi is a complex network analysis software based on JAVA program, used for analysis and visualization of various networks and complex systems. Community detection algorithm (also known as community discovery algorithm) is a technology used to reveal network aggregation behavior. We can treat each single herb as a node, and regard the prescription as a kind of community of nodes with the same characteristics. The closer the relationship between the herbs in the prescription, the easier they are to be divided into the same community. The higher the degree value of the node, the greater the importance of the medicinal herb, and the closer it is to the CP.

2.3. Chemical composition and targets of the core prescription (CP)

This part of the research is based on the TCM research platform of Traditional Chinese Medicine integrated with pharmacology, TCMIP v2.0 (http://www.tcmip.cn/TCMIP/index.php/Home/Login/login.html) and HERB platform. TCMIP 2.0 was developed by the China Academy of Chinese Medical Sciences. It mainly includes the five major database resources of "Encyclopedia of Chinese Medicine", "Chinese Medicine Prescription Database", "Chinese Medicine Ingredient Database", "Chinese Medicine Target Database", and "Disease Related Molecular Database". HERB (Chinese Medicine Group Identification) was jointly developed by the China Academy of Chinese Medical Sciences and Beijing University of Chinese Medicine. It collected 17 886 TCM-related papers published since 2011 by PubMed text mining, and then manually extracted 1241 gene targets and 494 modern diseases for herbs/ingredients from 1966 unique references (137 references for herbs and 1860 references for ingredients, total 1997 records) among them. In this study, TCMIP and HERB were used to obtain the chemical composition and molecular targets of the CP.

2.4. Disease targets database building

Through the integration of multi-source databases, relevant targets of liver cancer were collected. The databases used in this study are the GeneCards database (http://www.genecards.org/) and the Online Mendelian Inheri-tance In Man (OMIM) database (http://www.omim.org/). In order to select targets that are highly related to diseases, we took the median of the score values of the targets in the GeneCards database 4 times to obtain the core targets, and then took the intersection of these targets and the targets obtained in the OMIM database for screening, relevant disease targets were obtained.

2.5. Functional annotation and enrichment analysis

Metascape (http://metascape.org/gp/index.html#/main/step1) was used to analyze the targets obtained, the species for selection was "Homo sapiens", and custom analysis mode was used, set P value < 0.01, min overlap > 3, KEGG pathway result was obtained. RStudio's ggplot2 was used to draw the remaining items into a bubble chart. Finally, the bubble graphs of each herb were combined.

2.6. Construction of the CP access network

In order to better explain the overall mechanism of the CP, all targets of the CP were summarized into one Subetwork by using the Kyoto Encyclopedoa of Genes and Genomes (Kyoto Encyclopedoa of Genes and Genomes, KEGG) online tool Mapper-Search & Color Pathway. The intermediate genes were hidden for better display.

2.7. Gene survival analysis of core herbs

RNAseq (FPKM) and survival data of Liver Cancer (LIHC) were downloaded from GDC and TCGA public cancer genomics databases through UCSC Xena (http://xena.ucsc.edu/), and the R language survival package was used to analyze the core herbs. Single factor cox regression analysis was performed on 176 genes. At the same time, we used Kaplan-Meier Plotter (https://kmplot.com/analysis/) to show the effect of different expression levels of some genes on survival.

3. RESULT

3.1. Patient characteristics

In this study, 101 qualified patients with complete information of their age, gender, course of disease, and prescription information were selected for a total of 200 consultations, with an average of 2 consultations per patient. Among them, Physician 1 had 50 patients and Physician 2 had 51 patients, with 100 consultations each. The male to female ratios for the two Physicians were 7.33: 1 and 3.64: 1, respectively. The average age of the patients are 55.9 and 56.6 years, respectively. The median duration of the patient's disease are 2 and 6 months, respectively, and the patient's course of disease is mainly concentrated between 1-2 months (Supplementary Figure 1).

3.2. Analysis of herbal characteristics

Among the 200 prescriptions, there are 271 kinds of Chinese herbs. The frequency of herbs is 2958 times in total. The frequency of cold and warm herbs is the highest, which are 1225 and 812 times. The frequency of mild herbs is 556. Cold herbs appeared 352 times, and hot herbs appeared 13 times. Physician 1 prefers cold medicine, and Physician 2 prefers cool medicine. The most frequent flavors of herbs used are bitter, sweet, and pungent, which are 1752, 1419 and 1059 times respectively, and less than 200 times were recorded for the rest flavors. Physician 1 prefers to use bitter herbs, and Physician 2 prefers to use bitter herbs and sweet herbs (Supplementary Figure 2, Supplementary Table 1).

Figure 2. The screening process of CP (core prescription) for liver cancer.

The purple in the picture is the herb community of Physician 1, the green is the herb community of Physician 2, and the red is the non-core herbs used by the two Physicians.

Table 1.

Top 15 pathways in KEGG enrichment by two physicians

KEGG	Pathway	Tyle	Physician
hsa04151	PI3K (phosphatidylinositol-3-kinase)-Akt signaling pathway	Signal transduction	1&2
hsa04066	HIF-1 (Hypoxia Inducible Factor 1) signaling pathway	Signal transduction	1&2
hsa04218	Cellular senescence	Cell growth and death	1&2
hsa04210	Apoptosis	Cell growth and death	1&2
hsa04657	IL-17 (Interleukin 17) signaling pathway	Immune system	1&2
hsa04668	TNF (Tumor Necrosis Factor) signaling pathway	Signal transduction	1&2
hsa04115	p53 (Tumor Protein 53) signaling pathway	Cell growth and death	1&2
hsa04659	Th17 cell differentiation	Immune system	1&2
hsa04926	Relaxin signaling pathway	Endocrine system	1&2
hsa04625	C-type lectin receptor signaling pathway	Immune system	1&2
hsa04380	Osteoclast differentiation	Development and regeneration	1&2
hsa04660	T cell receptor signaling pathway	Immune system	1&2
hsa04068	FoxO signaling pathway	Signal transduction	1
hsa04620	Toll-like receptor signaling pathway	Immune system	1
hsa04152	AMPK (Adenosine 5-monophosphate(AMP)-activated protein kinase) signaling pathway	Signal transduction	1
hsa04010	MAPK (mitogen-activated protein kinase) signaling pathway	Signal transduction	2
hsa04915	Estrogen signaling pathway	Endocrine system	2
hsa04919	Thyroid hormone signaling pathway	Endocrine system	2

3.3. Complex network analysis based on gephi

We used COOC3.9 to convert the matrix format of 200 prescription herbs, and imported the matrix file into gephi0.9.2, selected the matrix format, selected the type of graph to be undirected, selected Modularity, and adjusted the resolution to 0.8. The module coefficient is 0.266, the prescriptions were divided into 3 communities, and the Fruchterman Reingold form was selected to adjust the layout (Figure 2). The top 10 herbs in terms of degree are Ezhu (Rhizoma Curcumae Phaeocaulis), Baihuasheshecao (Herba Hedyotdis), Huangqi (Radix Astragali Mongolici), Banzhilian (Herba Scutellariae Barbatae), Chuanbeimu (Bulbus Fritillariae Cirrhosae), Gancao (Radix Glycyrrhizae), Doukou (Fructus Amomi Rotundus), Danshen (Radix Salviae Miltiorrhizae), Shuihonghuazi (Polygoni Orientalis Fructus), Jineijin (Endothelium Coreneum Gigeriae Galli). Among them, Huangqi (Radix Astragali Mongolici), Danshen (Radix Salviae Miltiorrhizae) and Shuihonghuazi (Polygoni Orientalis Fructus) belong to Physician 1’s community. Ezhu (Rhizoma Curcumae Phaeocaulis), Baihua-sheshecao (Herba Hedyotdis), Huangqi (Radix Astragali Mongolici), Banzhilian (Herba Scutellariae Barbatae), Chuanbeimu (Bulbus Fritillariae Cirrhosae), Doukou (Fructus Amomi Rotundus) and Jineijin (Endothelium Coreneum Gigeriae Galli) belong to Physician 2's community. And Licorice belongs to other non-core herb communities. Gancao (Radix Glycyrrhizae) is a regulative herb in TCM prescriptions with no distinct feature. Therefore, we have chosen top 3 herbs used by two physicians respectively in terms of degree ranking, thus formulating the CP of the two physicians for the treatment of liver cancer. There are 6 herbs in total, namely Ezhu (Rhizoma Curcumae Phaeocaulis), Baihuasheshecao (Herba Hedyotdis), Huangqi (Radix Astragali Mongolici), Banzhilian (Herba Scutellariae Barbatae), Danshen (Radix Salviae Miltiorrhizae) and Shuihonghuazi (Polygoni Orientalis Fructus).

3.4. Overall targets of the CP for liver cancer treatment

To reveal the mechanism of CP in liver cancer treatment, we collected and analyzed the targets of CP. From the HERB and TCMIP database, 6 herbal targets of CP were obtained, and the targets of the herb and the disease were intersected. The total of 176 interactive targets were obtained and displayed visually. (Figure 3). Among them, there are 38 targets of Ezhu (Rhizoma Curcumae Phaeocaulis), 20 targets of Baihuasheshecao (Herba Hedyotdis), 76 targets of Banzhilian (Herba Scutellariae Barbatae), 113 targets of Huangqi (Radix Astragali Mongolici), 108 targets of Danshen (Radix Salviae Miltiorrhizae), and 35 targets of Shuihonghuazi (Polygoni Orientalis Fructus).

Figure 3. Herb targets show.

The connection between different herbal targets: blue-orange represents Huangqi (Radix Astragali Mongolici), and yellow represents Shuihonghuazi (Polygoni Orien-talis Fructus), red in the outer circle represents Baihuasheshecao (Herba Hedyotdis), blue represents Banzhilian (Herba Scutellariae Barbatae), green represents Danshen (Radix Salviae Miltio-rrhizae), purple represents Ezhu (Rhizoma Curcumae Phaeocaulis). The purple curve connects the same genes. The inner circle represents the gene list. The same genes of multiple herbs are shown in dark orange. Genes that have only appeared once are shown in light orange.

3.5. Functional enrichment analysis of core herbs

Metascape was used to perform kegg enrichment of genes interacting with CP and diseases (Min Overlap = 3, P value Cutoff < 0.01), with a total of 112 pathways obtained. The top 5 important pathway categories are signal transduction, immune system, Endocrine system, Metabolism, Nervous system (Supplementary Table 2). KEGG analysis showed that there are 16 metabolism pathways (Figure 4). Huangqi (Radix Astragali Mongolici) was mainly responsible for the fatty acid metabolism pathway, while Danshen (Radix Salviae Miltiorrhizae) and Baihuasheshecao (Herba Hedyotdis) were mainly responsible for the carbon metabolism pathway. Endocrine system-related pathways were abundant. Compared with the energy metabolism pathway, the related pathways in the immune and endocrine systems of Shuihonghuazi (Polygoni Orien-talis Fructus) and Banzhilian (Herba Scutellariae Barbatae) significantly increased, while the Baihuashe-shecao (Herba Hedyotdis) showed a downward trend. In the CP, Shuihonghuazi (Polygoni Orientalis Fructus) and Banzhilian (Herba Scutellariae Barbatae) mainly played a role in the immune and endocrine systems, while Baihuasheshecao (Herba Hedyotdis) mainly played a role in energy metabolism. Huangqi (Radix Astragali Mongolici), Danshen (Radix Salviae Miltio-rrhizae), and Ezhu (Rhizoma Curcumae Phaeocaulis) were involved in a wide range of interventions.

Table 1.

Characteristics of top 20 herbs

Herb	Frequency	Properties	Flavor
Baihuasheshecao (Hedyotis Diffusa)	152	Cold	Bitter,Sweet
Ezhu (Curcumae Rhizoma)	139	Warm	Pungent,Bitter
Banzhilian (Scutellariae Barbatae Herba)	122	Cold	Pungent,Bitter
Doukou (Amomi Fructus Rotundus)	102	Warm	Pungent
Chuanbeimu (Fritillariae Cirrhosae Bulbus)	95	Cold	Bitter,Sweet
Jineijin (Galli Gigerii Endothelium Corneum)	89	Mild	Sweet
Chonglou (Paridis Rhizoma)	89	Cold	Bitter
Huangqi (Astragali Radix)	80	Warm	Sweet
Shancigu (Cremastrae Pseudobulbus Pleiones Pseudobulbus)	79	Cool	Sweet,Pungent
Jixingzi (Impatientis Semen)	77	Warm	Bitter,Pungent
Dannanxing (Arisaema Cum Bile)	72	Cool	Bitter,Pungent
Fengfang (Vespaenidus)	70	Mild	Sweet
Muhudie (Oroxyli Semen)	70	Cool	Bitter,Sweet
Visciherba (Hujisheng)	56	Mild	Bitter
Danshen (Salviae Miltiorrhizae Radix Et Rhizoma)	55	Cold	Bitter
Yujin (Curcumae Radix)	51	Cold	Pungent,Bitter
Shuihonghuazi (Polygoni Orientalis Fructus)	50	Cold	Salty
Kushen (Sophorae Flavescentis Radix)	46	Cold	Bitter
Rehmanniae Radix (Dihuang)	28	Cold	Sweet,Bitter
Suanzaoren (Ziziphi Spinosae Semen)	27	Mild	Sweet,Sour

Figure 4. KEGG enrichment analysis.

Bubble plot of KEGG enrichment pathways. Each bubble represents a KEGG pathway. The size of the bubbles is related to the relative ratio of targets to total targets on each pathway, with larger bubbles representing more abundant genes. −LogP shows the statistical significance of the P value. The larger the number, the more significant the P value. Establishing a network of pathways for the actions of core herbs

We selected the top 15 pathways with the most important influence in Table 1, omitted some target genes that lack upstream and downstream continuation, and summarized and displayed them (Supplementary figure 3). There were 12 important pathways commonly intervened by two physicians. There were 6 separate pathways with 3 for each physician.

3.6. Gene survival analysis of core herbs

Single factor cox regression analysis was performed on 176 genes of core herbs. Setting P < 0.01, a total of 14 genes significantly correlated with survival and prognosis were obtained. Among them, there are 2 genes with the Hazard Ratio (HR) between 0 and 1, which are considered to have protection for the prognosis of liver cancer; 12 genes with the HR greater than 1, these genes are considered to be risky for the prognosis of liver cancer (Supplementary Table 3). We selected the most representative gene (the smallest P value) from the two types of genes, Cytochrome C Oxidase I (MT-CO1) and Erythropoietin (EPO) and used Kaplan-Meier Plotter to show the effect of different expression levels on survival prognosis. Since there is no Mitochondrially Encoded MT-CO1 information in the database, we replaced it with Peroxisome Proliferator-Activated Receptor Gamma Coactivator 1-Alpha (PPARGC1A) in display Supplementary figure 4).

Table 2.

Top 5 Pathway Categories

KEGG PATHWAY	Frequency
Signal transduction	24
Immune system	18
Endocrine system	16
metabolism	16
Nervous system	9

4. DISCUSSION

"China Guidelines for The Primary Liver Cancer Diagnosis and Treatment Guidelines 2017"³² pointed out that TCM relieves clinical symptoms, improve the body's resistance, and reduce the adverse reactions of radiotherapy and chemotherapy. "China Guidelines for The Diagnosis and Treatment of Primary Liver Cancer 2019"³³ has high-level evidence that taking TCM granules after liver resection can reduce the recurrence rate and improve the survival rate, which proves that TCM is very helpful for the treatment of HCC. Clinical studies have found that patients with liver cancer can benefit from TCM treatment during the whole process of the disease starting from onset. TCM also helps to reduce the risk of postoperative recurrence with such advantages as relatively good compliance, safety and tolerance, which is consistent with the patients’ group information characteristics in our study.³⁴

The properties, flavors, meridian tropism, clinical efficacy and indications are the standards guiding the use of Chinese materia medica, and it is also the applying method for TCM theory. In this study, descriptive analysis was used to comprehensively analyze the four properties, five flavors, meridian tropism, and doses of Chinese materia medica. The properties, flavors and efficacy constitute the fundamental basis for the analysis of TCM and clinical application.³⁵ In general, the frequency of cold herbs is significantly higher than that of hot medicines, and the frequency of bitter herbs is higher than that of sweet and pungent herbs. Cold herbs have a heat-scavenging effect, while warm herbs are of complementary effect.^36,37 This finding reflects the TCM concept of replenishing deficient substances and removing excess substances from the body to regulate homeostasis.

A TCM prescription consists of a variety of herbs with different flavors. The core herbs in the prescription are used to treat diseases, and the addition and subtraction of herbs are often used to relieve the symptoms of patients. At present, the network community analysis method in machine learning is often used to obtain the CP in TCM prescriptions, and reveal the mechanisms.³⁸ The 200 TCM prescriptions finally yielded a CP containing 6 herbs through the network community analysis. This CP consists of contributions from two physicians. Physician 1 contributed sub-prescriptions of Huangqi (Radix Astragali Mongolici), Danshen (Radix Salviae Miltiorrhizae), and Shuihonghuazi (Polygoni Orientalis Fructus); Physician 2 contributed sub-prescriptions of Baihuasheshecao (Herba Hedyotdis), Banzhilian (Herba Scutellariae Barbatae), and Ezhu (Rhizoma Curcumae Phaeocaulis). Physician 1's prescription mainly used the strategies of invigorating Qi and removing pathogenic factors, promoting blood circulation and resolving blood stasis, and the Physician 2’s strategies mainly focused on clearing away heat and detoxification. These strategies have been proved to be effective therapies in the treatment of tumors.^22,39

Astragalus polysaccharides (AP) is the main phytochemical component extracted from the TCM herb Huangqi (Radix Astragali Mongolici). AP can promote the secretion of IL-2, IL-12 and TNF-α, and reduce the level of IL-10 in serum, which has anti-tumor activity by improving the immune response of the host body.^40,41 The tumor microenvironment (TME) determines the nature and properties of cancer tissues by secreting cytokines and other substances throughout the environment, which are critical for tumor progression and metastasis.⁴² Cancerous tissues in the micro-environment depend on angiogenesis to meet their nutrient and oxygen needs. As a result, angiogenesis also serves as a pathway for metastasis. The extracts of both Danshen (Miltiorrhizae Radix Et Rhizoma) and Shuihonghuazi (Polygoni Orientalis Fructus) have anti-inflammatory and anti-angiogenic activities.^43,⇓-45 These herbs increase vascular immune cell infiltration by enhancing vascular normalization and have a synergistic effect on immunotherapy.⁴⁶

The physician 2's sub-prescription used the strategy of clearing away heat and detoxification. The strategy of clearing away heat and detoxification is one of the principles of TCM treatment of cancer, balancing oxidative stress and detoxification, including regulating energy metabolism and reducing the levels of harmful metabolites, which have important implications for tumor cell proliferation and metastasis.⁴⁷ Tumor cells secrete inflammatory factors that promote inflammation in the tumor microenvironment. Inflammation also suppresses immune responses in cancer patients.⁴² The extracts of Baihuasheshecao (Herba Hedyotdis), Banzhilian (Herba Scutellariae Barbatae), and Ezhu (Rhizoma Curcumae Phaeocaulis) can inhibit the inflammatory response of tumors and induce apoptosis of tumor cells.^48,⇓-50 Therefore, our CP treated tumor through different strategies. These two strategies conform to the TCM theory and the mechanism of molecular biology.

The sub-prescriptions of the two physicians co-involved 12 pathways in the top 15 pathways enriched by KEGG, and most of these pathways were related to signal transduction and immune system. In our further aggregated subnetworks, we found that the CP mainly exerted regulation through insulin like growth factor 1 (IGF-1)/ mammalian target of rapamycin (mTOR) pathway (inhibition of IGF-1/mTOR pathway to mediate apoptosis), angiogenesis and metastasis (inhibition of hepatocyte angiogenesis and liver cancer), Pro-apoptotic proteins and pro-apoptotic proteins [stop the mRNA of pro-apoptotic markers such as Bcl-2 (Bcl-2 apoptosis regulator), CASPASE-3, and nuclear factor kappa B (NF-κB) to reduce cancer activity], cell cycle regulation (regulate cell cycle). Studies have found that astragalus polysaccharide (APs) promotes chemotherapeutic drugs-induced tumor cell apoptosis by down-regulating the expression of Bcl-2, increasing the expression of Bax and caspase 3, and activating the JNK1/2 signaling pathway.⁵¹ Danshen (Miltiorrhizae Radix Et Rhizoma) inhibits cell proliferation, invasion and migration and induces HCC-related apoptosis in vitro by regulating the PI3K/AKT pathway and EGFR (Epidermal Growth Factor Receptor).⁵² Shuihonghuazi (Polygoni Orientalis Fructus) improves the tumor microenvironment (TME) by enhancing the activity of liver antioxidant enzymes (SOD, GPx and GRd) and reducing the levels of TNF-α, IL-1β and IL-6. Active ingredients of Baihuasheshecao (Herba Hedyotdis) like kaempferol can inhibit the activity of PI3K/Akt, MAPK and VEGF pathways and down-regulate downstream inflammatory factors such as nitric oxide synthase (iNOS), cyclooxygenase 2 (COX-2), and inhibitor of NFκB kinase (IKK) to reduce the inflammatory response of tumors.⁵³ Total flavonoids of Scutellaria barbata inhibits the growth and invasion of hepatoma cells by reducing the expression of MMP and increasing the expression of TIMP.⁵⁴ The ethanolic extract of Ezhu (Rhizoma Curcumae Phaeocaulis) can induce apoptosis of AGS cells through Bcl-2 and Bax.⁵⁵ In addition, studies have found that factors related to the nervous system and endocrine system may be involved in regulating the migration, proliferation and apoptosis of liver cancer cells.^56,57 Chinese herbal medicine can also improve liver function and inhibit the proliferation and metastasis of liver cancer by regulating energy metabolism.^58,59 These all provide new ideas for the treatment of liver cancer with TCM.

We found 176 genes potentially related to liver cancer in CPs. Through single factor cox regression analysis, 14 genes are significantly associated with survival and prognosis of liver cancer (P < 0.01). For example, Peroxisome proliferator-activated receptor gamma coactivator-1 alpha (PPARGC1A, formerly known as PGC-1a) is a transcriptional coactivator and metabolic regulator that plays a crucial role in regulating mitochondrial biogenesis. The study found that PGC-1a mRNA levels were significantly reduced in HCC tissues. PGC-1a acts as a tumor suppressor in the formation and progression of hepatocellular carcinoma. PGC-1a overexpression in HCC cell lines promotes apoptosis and may be a potential therapeutic target for hepatocellular carcinoma.⁶⁰ EPO (erythropoietin) is a protein-coding gene. Many tumors express EPO and/or EPO receptors. Various tumors express EPO to improve angiogenesis, thereby facilitating tumor growth and spread.⁶¹ HCC Patients with EPO or EPO receptor expression had poorer overall survival compared with patients with EPO or EPO receptor-negative HCC. EPO may serve as a key molecular prognostic marker for HCC survival.⁶² These findings are exciting and may suggest that key genes regulating liver cancer treatment and prognosis can be more easily discovered by analyzing CP. This not only greatly improves the efficiency of discovering key Chinese herbs from large-scale case prescriptions, but also provides a direction for further research on the regulatory mechanism of core herbs.

5. LIMITATIONS AND INNOVATION

Great progress has been made in the studies of anti-hepatoma treatment with TCM, and some clinical studies have also achieved encouraging results.^63,64 However, the composition of TCM prescriptions is complex. Even though the TCM prescription works well for HCC, but the anti-cancer effect would be weakened or even disappear when a single herb or a single component is used. To some extent, the samples included in this study are insufficient. In addition, illustrious senior TCM physicians are engaged in medical treatment in different regions. Their patients' diet and living habits are different. The specific regulation of the CP on the upstream and downstream of the disease genes and the clearer effects of different gene expression levels on the prognosis still need further experimental verification.

TCM is a complex hybrid system with multiple components and multiple targets. Identifying potential targets and pathway mechanisms is the basic work. Machine learning, network pharmacology, and genomics have emerged as promising approaches to accelerating drug development, elucidating the mechanisms of action of TCM prescriptions, and understanding complex interactions between biological systems, drugs, and diseases.

6. CONCLUSIONS

Based on the liver cancer cases of two illustrious senior TCM physicians, we obtained the CP for liver cancer treatment through data mining. The CP may mainly exert anti-cancer effects by inhibiting inflammatory response, angiogenesis, and enhancing the body's immune response. We screened out 14 genes in the CP that may be related to the prognosis of liver cancer, and these genes may play an important regulatory role in the prognosis of liver cancer.

7. ACKNOWLEDGMENTS

The authors gratefully acknowledge the assistance of Yin Changjian (Liver Disease branch, Shandong Hospital of Traditional Chinese Medicine) and Dong Lei (School of Traditional Chinese Medicine, Beijing University of Chinese Medicine), and also present heartfelt thanks to Shao Yingjun (School of Humanities, Beijing University of Chinese Medicine) for language modification, and to all the patients for their participation.

Figure 1. Patient characteristics.

A. The patient’s age distribution, the black horizontal line in the figure represents the average age of the patient. B. The patient’s course of disease distribution.

Figure 2. Herbal characteristics.

A. Proportion of medicinal properties; B. Proportion of medicinal flavors; In graph C, the y-axis is the patient's serial number, and the x-axis is the herbs. The characteristics of herbs are represented by cold, cool, mild, warm and hot, and the size of the bubbles indicates the dosage. The larger the bubble, the greater the dose. The color corresponds to the x-axis attribute. In graph D, the Y-axis is the serial number of the patient, and the X-axis is the corresponding herbs to the figure In graph E. In graph E, The X-axis is the top 20 drugs in terms of frequency, which are arranged in groups of warm, mild, and cold, and the Y-axis is the frequency of the herbs.

Figure 3. The subnetwork shows the gene pathway mechanism and regulation of the CP of two illustrious senior TCM physicians.

Light grey lines represent cell membranes and gene names are shown as rectangles. Markers below the genes show targeting patterns of the two CPs. To better illustrate the research question, the genes in the middle are omitted and indicated by dashed arrows.

AKT(Serine/Threonine Kinase), AP1(Activator Protein 1), BAX(Bcl2 Associated X, Apoptosis Regulator), BCL2(B-Cell Lymphoma-2), CASP3/8/9(Caspase), c-Fos(Proto-Oncogene, Ap-1 Transcription Factor Subunit), COX2(Cyclooxygenase 2), CXCL8(C-X-C Motif Chemokine Ligand 8), EBAG(Estrogen Receptor Binding Site Associated Antigen), ENOS(Endothelial Nitric Oxide Synthase), ER(Estrogen Receptor), ERK(Extracellular Regulated Protein Kinases), GK(Glycerol Kinase), HIF-1α(Hypoxia Inducible Factor 1 Subunit Alpha), IGF1R(Insulin Like Growth Factor 1 Receptor), IKBα(Inhibitor Of Nuclear Factor), IL-1/6/8(Interleukin), INOS(Inducible Nitric Oxide Synthase), INSR(Insulin Receptor), Jak(Janus Kinase), LAP/XIAP(Laryngeal Adductor Paralysis/X-Linked Inhibitor Of Apoptosis), Mcl1(Apoptosis Regulator, Bcl2 Family Member), MDM2(Proto-Oncogene), MMP1/9(Matrix Metallopeptidase), NFκB(Nuclear Factor Kappa B Subunit), P21/53(Tumor Suppressor Gene), PAI(Plasminogen Activator Inhibitor), PARP(Poly(Adp-Ribose) Polymerase), PGR(Progesterone Receptor), PI3K(Phosphatidylinositol-3-Kinase), PPARγ(Peroxisome Proliferator Activated Receptor), PKB(Protein Kinase B), PKC(Protein Kinase C), PTEN(Phosphatase And Tensin Homolog), RARα(Retinoic Acid Receptor Alpha), RTK(Receptor Tyrosine Kinase), SIRT(Sirtuin), SYK(Spleen Associated Tyrosine Kinase), TIMP1(Metallopeptidase Inhibitor 1), TLR2/4(Toll Like Receptor2/4), TNF(Tumor Necrosis Factor), VEGF(Vascular Endothelial Growth Factor).

Table 3.

Gene statistics related to survival in core herbs（p<0.01）

Gene_Symbol	Gene_ID	coef	Hazard.Ratio(HR)	CI95	P.Value
MT-CO1	4512	-0.2663	0.7662	0.6388-0.9191	0.0041
PPARGC1A	10891	-0.1854	0.8308	0.7280-0.9481	0.0060
EPO	2056	0.2606	1.2977	1.1746-1.4337	0.0000
MMP1	4312	0.3369	1.4006	1.1914-1.6466	0.0000
HMOX1	3162	0.2239	1.2510	1.1098-1.4101	0.0002
CXCL8	3576	0.1620	1.1758	1.0780-1.2826	0.0003
UGT1A8	54576	0.4095	1.5061	1.1686-1.9411	0.0016
SERPINE1	5054	0.1267	1.1350	1.0468-1.2307	0.0021
CA9	768	0.1152	1.1221	1.0417-1.2088	0.0024
CDK4	1019	0.2561	1.2919	1.0813-1.5436	0.0048
LGALS3	3958	0.1413	1.1517	1.0440-1.2706	0.0048
HIF1A	3091	0.2069	1.2298	1.0594-1.4276	0.0066
ICAM1	3383	0.1625	1.1764	1.0455-1.3237	0.0069
IFNB1	3456	5.1632	174.7229	4.0641-7511.6843	0.0071