Clinical phenotype network: the underlying mechanism for personalized diagnosis and treatment of traditional Chinese medicine

Abstract

Traditional Chinese medicine (TCM) investigates the clinical diagnosis and treatment regularities in a typical schema of personalized medicine, which means that individualized patients with same diseases would obtain distinct diagnosis and optimal treatment from different TCM physicians. This principle has been recognized and adhered by TCM clinical practitioners for thousands of years. However, the underlying mechanisms of TCM personalized medicine are not fully investigated so far and remained unknown. This paper discusses framework of TCM personalized medicine in classic literatures and in real-world clinical settings, and investigates the underlying mechanisms of TCM personalized medicine from the perspectives of network medicine. Based on 246 well-designed outpatient records on insomnia, by evaluating the personal biases of manifestation observation and preferences of herb prescriptions, we noted significant similarities between each herb prescriptions and symptom similarities between each encounters. To investigate the underlying mechanisms of TCM personalized medicine, we constructed a clinical phenotype network (CPN), in which the clinical phenotype entities like symptoms and diagnoses are presented as nodes and the correlation between these entities as links. This CPN is used to investigate the promiscuous boundary of syndromes and the co-occurrence of symptoms. The small-world topological characteristics are noted in the CPN with high clustering structures, which provide insight on the rationality of TCM personalized diagnosis and treatment. The investigation on this network would help us to gain understanding on the underlying mechanism of TCM personalized medicine and would propose a new perspective for the refinement of the TCM individualized clinical skills.

Keywords

personalized medicine / complex network / clinical phenotype network / traditional Chinese medicine

Introduction

Variability among individuals is one of the main issues for medical research and clinical tasks [1], which actually calls for personalized medicine [2,3] to obtain appropriate treatment for the individual patient with different diseases [4–6]. Advances in human genome research and systems biology are opening the door to a new paradigm of systems medicine and personalized medicine with aim for developing individualized treatment based on their genetic profiles [7–9]. The interconnections of both genotypes and phenotypes to the relevant diseases [10–12] compose a complex network [13] that provides insight of the pathophysiology of diseases and the outcome of individualized treatment. Current disease taxonomies like ICD-10, which are primary based on symptoms, can incorporate more molecular characteristics in order to refine their systems for an enhanced healthcare. Therefore, investigation on the inter-relationships between clinical phenotypes (including symptoms and signs) and genotypes, and the inner-relationships between clinical phenotypes to build a medical knowledge network is important for disease understanding and practical clinical solutions [14]. A number of studies [15–19] demonstrated the underlying molecular mechanisms of disease comorbidity and the clinical phenotype similarities between diseases, which indicate that to some extent the regularities on clinical phenotypes of diseases would have correlations to the underlying molecular mechanisms. This suggests that elaborating the relationships between clinical phenotypes plays an important role to understand disease phenomena. However, compared with the complex interactions between various molecular entities like genes, proteins and metabolities that have a large number of related studies [13,20], the complex relationships between clinical phenotypes, in particular between symptoms, have not been systematically investigated.

As a well-established medical science, traditional Chinese medicine (TCM) has been recognized as a unique personalized medicine, which mainly investigates the patterns of symptoms and signs manifested on patients to conduct diagnoses and prescriptions. The approach of personalized medicine in TCM originated from the first classical TCM literature—Inner canon of Emperor Huang [21], which was compiled more than two thousand years ago. It illustrates that TCM physicians should prescribe different therapies for different patients who live in different geographic regions, which would contribute to different disease pathologies, even they have same disease manifestations. TCM personalized medicine developed a rather comprehensive practical clinical framework to obtain effective treatment for various disease conditions.

Furthermore, due to the availability of a large number of clinical treatment options, including Chinese herb prescriptions, acupuncture and other types of physical therapies, TCM practitioners could deliver various kinds of individualized treatments for patients in order for obtaining maximum effectiveness and least side effects.

However, owning to the richness of medical literatures and the complicated real-world clinical settings, TCM researchers have not developed efficient evidence-based framework for knowledge distilling to get the most effective therapies for specific patients. Currently even for a specific patient, different TCM physicians would have different diagnosis and offer different treatments [22,23]. This would raise issues and challenges to TCM personalized medicine, i.e., (1) What rationality underlies the mechanisms held in this physician-oriented personalized medicine; (2) How TCM physicians obtain the best individualized treatment for individual patients. In this paper, we present a discussion on the related topics, investigate the underlying mechanisms and offer perspectives on the future development of TCM personalized medicine.

Traditional conceptual meanings of TCM personalized medicine

Personalized medicine is the long-standing clinical principle of TCM established since thousands of years ago. Placing human in the context of social and natural system, TCM takes symptoms and signs as the main evidences for personalized diagnosis and treatment. To utilize this kind of highest-level clinical phenotypes, which have dynamic and personalized characteristics, TCM practitioners have developed flexible and powerful theories to master the subtle changes and complicated relationships between clinical phenotypes. Based on the clinical phenotypes (i.e., symptoms and signs) elaborately observed by physicians and reported by patients themselves, and the analogy of the rules behind human biological system and natural/social systems, TCM established several well-known diagnostic principles [24], such as syndrome differentiation of eight principles, syndrome differentiation of six channel’s theory and syndrome differentiation of zang-fu viscera, to differentiate the states of disease for individual patients. In the most famous theoretical literature, Inner canon of Huangdi, the conceptual meanings of personalized medicine were well illustrated, which expatiated that personalized medicine in TCM concerns constitutional, etiological and environmental characteristics of patients to prescribe right treatment for right patient. This means that two patients with similar clinical manifestations (or similar diseases) could obtain different treatments, based on the underlying individualized pathologies and nosogeneses in the context of the interaction between human and its environment.

Modern clinical meanings of TCM personalized medicine

In the contemporary TCM clinical settings, the disease conditions of patients are mainly generalized as modern diseases or disease categories like ICD-10. This is also called disease-syndrome integration (Bing-Zheng Jie He), which both considers TCM diagnostic result (i.e., syndrome) and modern diagnostic result (i.e., disease) for TCM personalized medicine [25]. That is, in current clinical settings, individualized treatments are prescribed for patients who have the same diseases or disease categories like type 2 diabetes. In this condition, syndrome types are the main evidences for defining appropriate individualized treatments. The current disease taxonomies like ICD-10, which mainly considering symptoms, microscopic examinations and laboratory test, have well recognized shortcomings including a lack of sensitivity in identifying preclinical disease and a lack of specificity in defining disease unequivocally [26]. The inclusion of molecular information would provide great enhancement to disease classification refinement and personalized medicine [14]. On the other hand, TCM personalized medicine delivers another kind of successful approach, which focuses on incidental patient characteristics, socio-environmental influences on diseases and in particular on the interactions between these factors with symptoms that makes TCM achieve personalized medicine without inclusion of molecular mechanisms.

Contemporary clinical diagnosis comprises the personalized knowledge of TCM practitioners

In real-world TCM clinical settings, like inpatient encounters and outpatient encounters, various TCM diagnoses and treatments are made for different patients who have same modern diseases like type 2 diabetes, stroke and heart diseases. The clinical diagnosis of patient to syndromes is an epistemological process [27], which accordingly would be physician oriented and biased. This means that even for the same patient he/she would obtain different diagnosis results from different TCM physicians [22]. Furthermore, the diagnosis approach used would consciously influence the physicians for capturing of symptoms and signs. This has been confirmed in our data analysis on symptom similarity of 82 patients observed at the same time by three physicians (Fig.1). Using the cosine similarity method [28], we calculated the similarities between the symptoms recorded by these three physicians for each patient and also between the encounters held by each physician. It is interesting that the similarities between the symptoms captured by three different physicians on same patients, are most in low values (their mean is 0.26), i.e., 82.2% of the similarities are between 0.2 and 0.4. This means that only about 20% to 40% symptoms were applied as common symptom terms among these three physicians for same patients. This is really unexpected because in principle we would hope that much higher similarities should be found for same patients. Meanwhile, we also notice that there is no single symptom in common for about 22.4% pairs of patients. This result has a significant departure from we thought. This is caused by that for the same patient there are different manifestations, and the TCM physicians have their own preference of selecting the use of symptoms.

To check whether the symptom similarity of the encounters of one physician would tend to be higher than expected, we calculated the symptom similarities between each encounters of these three physicians. The results show that, although the mean value of symptom similarities between different patients by the same physician is less than that between different physicians on same patients (physician 1: 0.16; physician 2: 0.24; physician 3:0.21; 3 physicians on same patients: 0.26), they still have some high similarity between 0.5 and 0.7. We further compared the symptom similarities of one patient to the similarities between the symptoms of this patient and the symptoms of other patients captured by the same physician. We found that there are a number of cases with higher similarities between different patients of same physician than the symptom similarity between different physicians on same patients (Table 1). These results indicate that besides of the personal bias of diagnosis, there exist strong selection biases for symptom capturing by different physicians. This means that the knowledge of physicians essentially influences their attention on different kinds of phenotypes and thus would have selection biases on symptoms manifested on patients.

This kind of bias in symptom selection would partially contribute to the personalized diagnosis of different TCM physicians for the same patient. However, multiple individualized diagnosis for the same patient conflicts with the diagnostic criterion of diseases. It is evidenced that more systematic comparison studies are needed to obtain the best diagnosis for a single patient. In this manuscript, we would try to explain its rationality from the perspectives of network medicine and show the appropriate way to improve it.

Various kinds of combination therapies would contribute to the TCM personalized medicine

Another aspect of TCM personalized medicine comes from the combination of various kinds of therapies like herbal medicine, acupuncture and naprapathy. The diversity of personalized therapy may come from the large number of herbs (i.e., thousands of different herbs) that could be used in combinatorial form in clinical prescriptions. The combinational herb therapies may possess similar efficacy due to the complicated interactions between herb ingredients and possible similar efficacy shared by different herb ingredients [29]. Therefore, we have seen various therapies used in contemporary TCM clinical practice for similar disease conditions. Actually, due to the lack of clinical evaluation studies and the empirical preferences of physicians, there would be more individualized therapies than the practical requirements of clinical practice in TCM field.

To illustrate this kind of personalized medicine phenomenon, we use the herb prescription data from a well-designed observational clinical study [22], which involved three highly-experienced TCM physicians to order herb prescriptions for 33 insomnia patients. In total, there are 99 herb prescriptions ordered for the 33 patients by the three physicians. We calculated the frequency of herb ingredients used in these three physicians. Fig. 2 depicts three groups of top 20 herbs used by these three physicians. These three herb groups include 37 distinct herbs and 17 herbs of which are used by only one physician but are not used by others. It shows that these three physicians prescribed very different herbs for the same insomnia patients. For example, the herbs like Dragon bone, Oyster shell, Spine date seed and Thinleaf milkwort root are used by physician 1 in all the cases while these herbs are rarely used by physician 3. On the other hand, Tuber fleeceflower stem is frequently used by physician 3 but it is less likely used by physician 1. Some herbs like Fermented soybean, Lophatherum herb, Lotus plumule, Dragon’s teeth, Chinese magnoliavine fruit and Red peony root, which are frequently used by physicians 2 and 3, are seldom used by physician 1. This result proposes a demonstration of the personalized herb treatment preferred by different TCM physicians for same patients. To further investigate the preferences of TCM practitioners for herb prescription, we used the corresponding 246 herb prescriptions treated by three physicians simultaneity for 82 patients to calculate the herb similarity between each prescription for same patients (Fig. 3). The results show that most herb similarities of each prescription for same patients are in the range of [0.2,0.4] and the corresponding mean of similarity is 0.27, which means that there are significant differences of herb ingredients between the prescriptions for same patients treated by three different physicians. However, different prescriptions treated by each physician for different patients hold a high degree of similarity. The mean herb similarity is 0.41. Particularly for physician 1, it has a much larger mean similarity of 0.62. This confirms the personal preference of herb usage in prescriptions by TCM physicians, which obviously contributes to the personalized treatment in TCM.

Clinical phenotype network constitutes the underlying mechanisms of TCM personalized medicine

Individualized therapies and the related personalized diagnoses have been widely recognized and considered as one of the main characteristics of TCM clinical schema. Together with the genetic profile based modern personalized medicine, it will be a real opportunity for researchers to develop appropriate personalized medicine for real-world patients with the help of practical principles of TCM personalized medicine because it has long period of clinical practice that makes its most related key steps evaluated and optimized by real-world clinical requirements. However, there are still several key aspects of TCM personalized medicine remain unclear in order for achieving optimal clinical effectiveness, such as the underlying mechanisms of personalized diagnosis, herb-clinical phenotype correspondence. Network medicine has become a promising approach for understanding the complicated phenomena of human diseases [13]. The future precision medicine is based on the integration of clinical phenotypes, genomics and molecular phenotypes to generate a medical knowledge network [30].

Relationships among clinical phenotype entities composing a clinical phenotype network

Clinical phenotypes like diseases and symptoms are complicated and usually co-occurred, which has the underlying molecular mechanisms in common [17,31]. The clinical phenotypes are the only materials used for diagnosis and treatment in TCM, the regularities held for clinical phenotypes (i.e., symptoms and signs) would thus contribute as main factors to the principle of TCM personalized diagnosis and treatment. Therefore, instead of focusing on capturing the specific manifestations of disease as modern biomedical physicians do, TCM practitioners apply systematic observation skills to detect the symptoms and signs of patients, and generated elaborate descriptions of symptom manifestations of patients [32]. These symptom entities are complicated related and co-occurred [33], which could be used for classifications of patients [34] and for personalized diagnosis and treatment [35]. Using the symptom features of insomnia data set from two related studies [22,36], we calculated the co-occurrence of symptoms on patients and constructed a network with symptoms as nodes and co-occurrences of symptoms as links (Fig. 2). To eliminate the influence of the manifestations of insomnia, we filtered out the insomnia related symptoms such as difficult to sleep, easy to awake and dreaminess, and pulse related signs from the data set. We connected two symptoms if they occurred for one patient. The obtained symptom network involved 442 nodes and 5671 links. It is noted that the average shortest path length of the symptom network is 2.16 and its diameter is 3, thus it has small-world structure in which most symptoms are closely related to each other. The nodes with top 5 high degree are dysphoria (222), dizziness (205), red tongue(197), thin yellow fur(163) and irritable(157), which are frequent occurred symptoms in insomnia as well. Furthermore, the average clustering coefficient of this network is 0.789. This is much larger than a random network which indicates that most symptoms are clustered together in dense groups. This offers insightful information for TCM personalized diagnosis: the symptoms on patients detected by one physician may conceive hidden related symptoms, which would be emphasized by another physician for personalized diagnosis (Fig. 1). Therefore, TCM physicians could acquire different diagnoses for one patient based on the clinical phenotypes they observed. However, owning to the possible molecular correlation between clinical phenotypes, their prescribed personalized treatment may obtain similar efficacies in such a case that no strict evaluation criterions are performed.

Strong correlations between herbs and phenotype entities

In TCM clinical settings, symptoms are the clinical phenotypes as direct evidences for clinical diagnosis and the targets for herbs. Strong correlations are noted between herbs and symptoms. In fact, the principle of formula-syndrome (Fang-Zheng) correspondence [37], which is advocated to organize clinical prescriptions for disease treatment, has demonstrated the evidence of correlations between symptoms and herbs because herb prescription consists of herb ingredients and syndromes are differentiated based on symptom manifestations. Several data mining studies have been conducted to detect the herb-symptom relationships [38]. For example, Poon et al. [39] applied biclustering methods to find the meaningful herb-symptom relationships from clinical data. A recent research [28] investigated the correlation between herbs and symptoms using four clinical data collections and found that there exist strong correlations between herbs and symptoms. Furthermore, it showed that the specific herb-symptom associations could be detected by complex network based approach. In fact, symptom-herb correlation is one of the basic principles used in TCM personalized treatment. In the context of clinical formula organization framework, specific herbs would be additionally ordered for patients according to their individualized symptoms. Likewise, TCM practitioners would prefer capturing some specific symptoms on patients if they prefer using some specific herbs. This would partially interpret the preference in symptom selection, which we have discussed above in this manuscript.

Syndrome network with shared phenotypes deciphers the rationality of physician-oriented diagnosis

Previous study [22] indicated that TCM physicians would have high probability to make different syndrome diagnoses for one single patient. Is there any rationality relevant to this obviously improper diagnosis phenomenon? In TCM clinical practice, syndrome co-morbidity is a popular phenomenon [33,40], which could partially contribute to different diagnoses when different TCM physicians have selection preference. To further investigate the promiscuous boundary between syndromes, we generated a syndrome network in which nodes represent syndromes and links represent shared symptoms using the insomnia data set. We excluded the signs of pulse and insomnia direct related symptoms like difficult to sleep, easy to awake, etc. because these features would be high possibility in common in different encounters. The resulted network (Fig. 5) displays an almost complete graph (94.8% links are connected) with most syndromes linking to all the other syndromes. There are 36 different syndrome types in the network, in which Phlegm heat syndrome, Liver stagnation syndrome and Spleen deficiency connected to all the left 35 syndromes. Furthermore, there are other 29 (80% in the 36 nodes) syndromes connecting to 34 syndromes. This indicates that syndromes actually have high probability to relate to each other by shared symptoms. Therefore, although different physicians make obviously different syndrome diagnoses for patient, they actually described the similar phenotypes of patients by highly related syndromes. Accordingly, we argue that it is the promiscuous boundary between syndromes (i.e., mixed symptoms and pathologies between syndromes) that sustains the existence of high degree of personalized diagnosis in TCM clinical settings. Further investigation of the underlying mechanisms of syndrome network and symptom interactions would be a significant task for TCM researchers to normalize the diagnostic procedure and result of different physicians.

Discussion

The fundamental idea behind personalized medicine in modern biomedicine is coupling the established clinical-pathological indexes with molecular profiling to create diagnostic, prognostic and therapeutic strategies precisely tailored to each patient’s requirements [30]. A medical knowledge network that integrates the intra and inter relationships between various levels of biomedical entities such as biological, physiological, pathological and environmental entities should be developed and refined to harbor basic biological discovery and clinical innovation [14]. It is the requirement of TCM as well to integrate the state of the art of molecular profiles [41–43] to improve the diagnostic and therapeutic capabilities of TCM personalized medicine. However, as we have demonstrated above, there is still an important task which was overlooked by most TCM researchers, and the task is investigating the mechanisms underlying clinical phenotype network in TCM to decipher the patterns behind personalized diagnosis and treatment.

Furthermore, the description granularity of clinical phenotypes in TCM clinical settings is much more elaborated and comprehensive than modern biomedicine [44]. Therefore, together with the established clinical oriented theories, TCM researchers hold more facilities for the discovery of medical insights hidden in the complicated clinical phenotype network. However, the preparation of high quality data on clinical phenotypes, symptoms and signs in particular, and socio-environmental factors, would be a critical step. Moreover, the correlation and regularity between clinical phenotypes and the underlying molecular phenotypes should be further investigated by integrated analysis of large-scale clinical and genomics data. It is a great opportunity for TCM to investigate the underlying mechanisms of their theoretical aspects and clinical regularities if fully plumbing this unexplored integrated clinical phenotype network. Specifically, to fully exert the potential of TCM personalized medicine, we need have improvement on the following three aspects in future work.

First, large-scale high quality clinical database should be curated to offer valuable resources for acquiring the ontological relationships between clinical phenotypes, syndromes and herbs, considering the epistemological consequence of TCM practitioners. When large-scale data are available, we could confirm ontological knowledge from the clinical data. In recent years, a medium-sized clinical data warehouse [45] has been developed to support medical knowledge discovery in TCM field, which has integrated about 300 000 clinical encounters from TCM hospitals and departments. TCM clinical records should be integrated and several real-world clinical data processing issues [44] should be addressed to provide an appropriate high quality big data platform for scientific research and translational medicine.

Secondly, through classical literature curation and data analysis, standard syndrome nomenclature should be proposed to develop a new syndrome classification system (SCS) for TCM clinical diagnosis. Currently, no well-accepted SCS is available for TCM clinical diagnosis. Therefore, TCM diagnoses are not well coded and standardized. The new SCS should contain not only conceptual definitions of syndromes but also rules to organize a clinical syndrome names with appropriate granularities. Clinical guidelines for TCM diagnosis should be proposed to normalize the diagnostic procedure based on the current best available evidences. These two tasks could obtain the support from the phenotype network data analysis using the big clinical data.

Thirdly, to develop a persistent refinement framework on personalized medicine, the inter relationships between symptoms, syndromes and therapies should be further plumbed and evaluated by clinical effectiveness. It is only clinical meaningful to develop personalized medicine when it could offer appropriate clinical solutions with maximized effectiveness for disease treatment [46]. Currently, the key obstacle for TCM personalized medicine is the lacking of high quality information of outcomes and treatment effectiveness. As a clinical based medicine, TCM hold the outcome related information in regular clinical encounters, which are mainly contained in electronic medical record (EMR). But EMR data are not enough to fully evaluate the effectiveness of TCM treatment, as it lacks of long-term outcomes. Therefore, it is urgent to acquire and record the clinical data for long-term outcome (e.g., during daily activities) after regular clinical encounters and build large-scale data centers to provide the services of data integration, processing and analysis.

References

[1] Piquette-Miller M, Grant DM. The art and science of personalized medicine. Clin Pharmacol Ther2007; 81(3): 311–315

[2] Lesko LJ. Personalized medicine: elusive dream or imminent reality? Clin Pharmacol Ther2007; 81(6): 807–816

[3] Hamburg MA, Collins FS. The path to personalized medicine. N Engl J Med2010; 363(4): 301–304

[4] Meyers DA, Bleecker ER, Holloway JW, Holgate ST. Asthma genetics and personalised medicine. Lancet Respir Med2014; 2(5): 405–415

[5] Mosli MH, Sandborn WJ, Kim RB, Khanna R, Al-Judaibi B, Feagan BG. Toward a personalized medicine approach to the management of inflammatory bowel disease. Am J Gastroenterol2014; 109(7): 994–1004

[6] Hutchinson L. Personalized cancer medicine: era of promise and progress. Nat Rev Clin Oncol2011; 8(3): 121

[7] Ginsburg GS, McCarthy JJ. Personalized medicine: revolutionizing drug discovery and patient care. Trends Biotechnol2001; 19(12): 491–496

[8] Pirmohamed M. Personalized pharmacogenomics: predicting efficacy and adverse drug reactions. Annu Rev Genomics Hum Genet2014 May 29. [Epub ahead of print]

[9] Weston AD, Hood L. Systems biology, proteomics, and the future of health care: toward predictive, preventative, and personalized medicine. J Proteome Res2004; 3(2): 179–196

[10] Chung KF. Defining phenotypes in asthma: a step towards personalized medicine. Drugs2014; 74(7): 719–728

[11] Raciti GA, Nigro C, Longo M, Parrillo L, Miele C, Formisano P, Béguinot F. Personalized medicine and type 2 diabetes: lesson from epigenetics. Epigenomics2014; 6(2): 229–238

[12] Fraser M, Berlin A, Bristow RG, van der Kwast T. Genomic, pathological, and clinical heterogeneity as drivers of personalized medicine in prostate cancer. Urol Oncol2014 Apr 22. [Epub ahead of print] doi: 10.1016/j.urolonc.2013.10.020

[13] Barabási AL, Gulbahce N, Loscalzo J. Network medicine: a network-based approach to human disease. Nat Rev Genet2011; 12(1): 56–68

[14] Committee on A Framework for Developing A New Taxonomy of Disease. Towards precision medicine: building a knowledge network for biomedical research and a new taxonomy of disease. Washington, DC: National Academies Press, 2011

[15] Rzhetsky A, Wajngurt D, Park N, Zheng T. Probing genetic overlap among complex human phenotypes. Proc Natl Acad Sci USA2007; 104(28): 11694–11699

[16] Blair DR, Lyttle CS, Mortensen JM, Bearden CF, Jensen AB, Khiabanian H, Melamed R, Rabadan R, Bernstam EV, Brunak S, Jensen LJ, Nicolae D, Shah NH, Grossman RL, Cox NJ, White KP, Rzhetsky A. A nondegenerate code of deleterious variants in Mendelian loci contributes to complex disease risk. Cell2013; 155(1): 70–80

[17] Lee DS, Park J, Kay KA, Christakis NA, Oltvai ZN, Barabási AL. The implications of human metabolic network topology for disease comorbidity. Proc Natl Acad Sci USA2008; 105(29): 9880–9885

[18] van Driel MA, Bruggeman J, Vriend G, Brunner HG, Leunissen JA. A text-mining analysis of the human phenome. Eur J Hum Genet2006; 14(5): 535–542

[19] Zhou X, Menche J, Barabási AL, Sharma A. Human symptoms-disease network. Nat Commun2014; 5: 4212

[20] Vidal M, Cusick ME, Barabási AL. Interactome networks and human disease. Cell2011; 144(6): 986–998

[21] Anonymous. The Inner Canon of Emperor Huang, Beijing: Chinese Medical Ancient Books Publishing House, 2003

[22] Jiang L, Liu B, Xie Q, Yang S, He L, Zhang R, Yan S, Zhou X, Liu J. Investigation into the influence of physician for treatment based on syndrome differentiation. Evid Based Complement Alternat Med2013; 2013: 587234

[23] Hu J, Liu B. The basic theory, diagnostic, and therapeutic system of traditional Chinese medicine and the challenges they bring to statistics. Stat Med2012; 31(7): 602–605

[24] Lu AP, Jia HW, Xiao C, Lu QP. Theory of traditional Chinese medicine and therapeutic method of diseases. World J Gastroenterol2004; 10(13): 1854–1856

[25] Xu H, Chen K. Integrative medicine: the experience from China. J Altern Complement Med2008; 14(1): 3–7

[26] Loscalzo J, Kohane I, Barabasi AL. Human disease classification in the postgenomic era: a complex systems approach to human pathobiology. Mol Syst Biol2007; 3: 124

[27] Liu B, Wang Y. Investigation on the concepts and their relationships among disease, symptoms and syndromes. J Tradit Chin Med2007; 48: 293–298

[28] Li Y, Zhou X, Zhang R, Wang Y, Peng Y, Hu J, Xie Q, Xue Y, Xu L, Liu X, Liu B. Complex network approach for investigating the herb-symptom correspondence phenomenon and analyzing the TCM clinical herb-symptom association knowledge. BMC Complement Altern Med2014 (Accepted)

[29] Zhou XZ, Zhang RS, Shah J, Rajgor D, Wang YH, Pietrobon R, Liu BY, Chen J, Zhu JG, Gao RL. Patterns of herbal combination for the treatment of insomnia commonly employed by highly experienced Chinese medicine physicians. Chin J Integr Med2011; 17(9): 655–662

[30] Mirnezami R, Nicholson J, Darzi A. Preparing for precision medicine. N Engl J Med2012; 366(6): 489–491

[31] O’Connor TG, McGuire S, Reiss D, Hetherington EM, Plomin R. Co-occurrence of depressive symptoms and antisocial behavior in adolescence: a common genetic liability. J Abnorm Psychol1998; 107(1): 27–37

[32] Zhou X, Peng Y, Liu B. Text mining for traditional Chinese medical knowledge discovery: a survey. J Biomed Inform2010; 43(4): 650–660

[33] Chen J, Lu P, Zuo X, Shi Q, Zhao H, Luo L, Yi J, Zheng C, Yang Y, Wang W. Clinical data mining of phenotypic network in angina pectoris of coronary heart disease. Evid Based Complement Alternat Med2012; 2012: 546230

[34] Zhao Y, Zhang NL, Wang T, Wang Q. Discovering symptom co-occurrence patterns from 604 cases of depressive patient data using latent tree models. J Altern Complement Med 2014; 20(4): 265–271

[35] Liu B, Chen S, Zhou X, Ni Q, He L. The principle of patient classification for type 2 diabetes based on symptoms. Beijing J Tradit Chin Med (Beijing Zhong Yi)2009; 28: 267–269 (in Chinese)

[36] Yan S, Zhang R, Zhou X, Li P, He L, Liu B. Exploring effective core drug patterns in primary insomnia treatment with Chinese herbal medicine: study protocol for a randomized controlled trial. Trials2013; 14(1): 61

[37] Zhang L, Wang, J, Wang Y. Research on FANG-ZHENG Correspondence. China J Tradit Chin Med Pharm (Zhonghua Zhong Yi Yao Za Zhi)2005; 20: 8–10 (in Chinese)

[38] Zhang XP, Zhou XZ, Huang HK, Feng Q, Chen SB, Liu BY. Topic model for Chinese medicine diagnosis and prescription regularities analysis: case on diabetes. Chin J Integr Med2011; 17(4): 307–313

[39] Poon J, Luo Z, Zhang RS. Feature representation in the biclustering of symptom-herb relationship in Chinese medicine. Chin J Integr Med2011; 17(9): 663–668

[40] Liu GP, Li GZ, Wang YL, Wang YQ. Modelling of inquiry diagnosis for coronary heart disease in Traditional Chinese Medicine by using multi-label learning. BMC Complement Altern Med2010; 10(1): 37

[41] Zhou X, Liu B, Wu Z, Feng Y. Integrative mining of traditional Chinese medicine literature and MEDLINE for functional gene networks. Artif Intell Med2007; 41(2): 87–104

[42] Su SB, Jia W, Lu A, Li S. Evidence-Based ZHENG: A Traditional Chinese Medicine Syndrome 2013. Evid Based Complement Alternat Med2014; 2014: 484201

[43] Li S, Fan TP, Jia W, Lu A, Zhang W. Network pharmacology in traditional chinese medicine. Evid Based Complement Alternat Med2014; 2014: 138460

[44] Liu B, Zhou X, Wang Y, Hu J, He L, Zhang R, Chen S, Guo Y. Data processing and analysis in real-world traditional Chinese medicine clinical data: challenges and approaches. Stat Med2012; 31(7): 653–660

[45] Zhou X, Chen S, Liu B, Zhang R, Wang Y, Li P, Guo Y, Zhang H, Gao Z, Yan X. Development of traditional Chinese medicine clinical data warehouse for medical knowledge discovery and decision support. Artif Intell Med2010; 48(2–3): 139–152

[46] Liu B, Zhang Y, Hu J, He L, Zhou X. Thinking and practice of accelerating transformation of traditional Chinese medicine from experience medicine to evidence-based medicine. Front Med2011; 5(2): 163–170

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg