Plasma cell mastitis (PCM) represents a serious inflammatory condition of breast that occurs in young and middle-aged females at nonpregnant and nonlactating period (Cutler, 1949). The main histopathological characteristics of PCM are the infiltration of plasma cells and lymphocytes in breast tissue (Yu et al., 2012). Interestingly, PCM shares similarity with breast cancer from the perspective of macroscopical or microscopical characteristics (Cheng et al., 2015). On the other hand, mounting evidence suggests that the dysregulation of immune system is closely associated with the pathogenesis of PCM. Recently, the incidence rate of PCM is quickly rising, yet the treatment methods for PCM are still rather limited. In clinical practice, surgical resection and hormone therapy remain two major treatments for PCM. Unfortunately, neither surgical resection nor hormone therapy could prevent recurrence of PCM. Moreover, the serious side effects of hormone therapy are still problematic (Fleming et al., 2022). Therefore, the discovery of effective PCM treatment or therapeutics with minimal side effects is clearly warranted.

Traditional Chinese Medicine (TCM) has a rather long history for the prevention and treatment of complex diseases in eastern Asia (Tu, 2011; Tu, 2016; Yin et al., 2022). Moreover, for decades, TCM has been often used as alternative or complementary medicine in the West. Indeed, Chinese herbal compounds have been successfully applied in the treatment of PCM in conjunction with Western medicine (Zhang et al., 2020a). In clinical practice, TCM refers to herbal entity prescription or formulae (also called ‘Fangji’), which may exhibit coordinating or synergistic effects through the combination of multiple herb drugs (Li et al., 2010). However, the design of formulae in TCM is solely based on the principle of ‘syndrome differentiation’ according to the medicinal properties of herbal entities. Moreover, the molecular mechanisms for the ‘formulae’ in TCM remain rather elusive.

Knowledge graph has emerged as an advanced technology in the field of artificial intelligence that is able to connect entities in a graph based on their existing intricate relationships (Ye et al., 2021). In particular, knowledge graph can enable the rational design and identification of combination therapies for a specific disease or phenotypes (Zeng et al., 2022). Recently, we developed and constructed a knowledge graph for the discovery of herbal drug combination toward immunotherapy and systematic immunity. Subsequently, we identified a synergistic combination of herbal drugs for PCM via a scoring system based on target-hitting rates and phenotype relativeness. To verify our concept of design, we conducted a clinical trial experiment for the drug combination of herbal compounds mentioned above (ClinicalTrials.gov registration: NCT05530226). Strikingly, our clinical results demonstrated that the herbal drug combination identified by knowledge graph can markedly suppress various inflammatory cytokines in serum, restore clinical symptoms, and reduce the recurrence rates of PCM patients with improved global health status.

In a previous study, we collected and compiled 240 targets for immunotherapy and systematic immunity from literature data (Zhang et al., 2020b). Recently, we collected 345 entities of herbal drugs documented in TCM books and herbal drugs announced by the National Administration of Traditional Chinese Medicine through advanced text-mining techniques. The existing intricate relationships between the herbal drugs and immunotherapy targets were also extracted and compiled via advanced text-mining techniques and manual curation for the construction of knowledge graph. We defined an ontology list consisting of 13 ontology terms describing the relations (edges) between herbal drug entities and the immunotherapy targets based on manual curation of literature data. Moreover, we collected the attributes of the medicinal properties of each herbal compound from the Pharmacopoeia of China (Hao and Jiang, 2015). We totally compiled and integrated 64 attributes of the medicinal properties of herbal drug entities into the knowledge graph. These medicinal properties are useful throughout the design of herbal drug combination. Finally, we built the knowledge graph using Neo4j and Py2Neo tools that consists of 895 nodes and 2197 edges (Figure 1, Figure 1—figure supplement 1), which can be visited online (http://www.ikgg.org/). Subsequently, we employed a scoring system (or so-called recommendation system) to asses and predict synergistic herbal drug combination from the knowledge graph. Note that the scoring function is able to identify those herbal drug combinations that are most related to specific phenotypes as well as herbal drug combinations that are able to hit most discrete cellular targets, yet still following the principle of ‘syndrome differentiation’ as described in the Pharmacopoeia of China (‘Methods’). Here, syndrome differentiation refers to the very basic principle of identifying and treating diseases in TCM. Syndrome (Zheng) is the presentation of the pathological changes during a specific disease course, including the location, cause, and nature of a disease. Moreover, ‘Jun-Chen-Zuo-Shi’ refers to the rules guided by syndrome differentiation to select multiple herbal drug entities to treat a specific disease in TCM. To this end, we used this scoring function to select herbal drug combinations consisting of eight herbal entities. We chose to identify drug combinations with eight entities because ‘formulae’ consisting of eight drugs are regarded as ‘essence combination’ in the TCM community. In short, we employed a combination generator that is able to randomly generate drug combinations with eight herbal drug entities for 10 rounds, each of which consists of 1000 random drug combinations. All the generated drug combinations from the 10 rounds were further ranked and evaluated. It is noteworthy that the scoring results of the 10 rounds are presented as normal distributions (Figure 2—figure supplement 1). The top 20 combinations from each round ranked by the scoring function were further curated and inspected by experts in TCM. Remarkably, we identified a specific drug combination that was ranked among the top 20 choices in all 10 rounds of calculation. The drug combination was chosen for further clinical study for two reasons. First, this drug combination was among the top 20 combinations in each round of our calculations. Second, we asked experts in TCM to inspect the top 20 combinations on the basis of ‘syndrome differentiation’ as described in the Pharmacopoeia of China, and finally the combination consisting of eight herbal drug entities, including ‘Fructus forsythiae,’ ‘Herba violae,’ ‘Uniflower swisscentaury root,’ ‘Danshen,’ ‘Astragalus,’ ‘Taraxacum,’ ‘Liquorice,’ and ‘Honeysuckle,’ was selected for further clinical study.

Figure 1 with 2 supplements see all

Download asset Open asset

Next, we extracted the subgraph for the herbal drug combination mentioned above and created a network diagram for the drug combination using Cytoscape tools (Reimand et al., 2019; Figure 2). In total, the eight herbal drug entities in the combination regulate 46 cellular targets related to immunotherapy and systematic immunity such as HIF-1 (Nan et al., 2022), iNOS, IL-17, IL-6, IL-1β, mTOR, NLPR3, PD-L1, STAT3 (Liu et al., 2020), TGF-β, TLR2, and TLR4 (Figure 2). It is noteworthy that the medicinal properties of the eight drug entities could be classified into three major categories of ‘heat-clearing and detoxicating,’ ‘Qi-tonifying,’ and ‘blood-activating menstruation regulating.’ Moreover, we conducted pathway analysis for the herbal drug combination for plasma cell mastitis. Interestingly, we revealed that the herbal drug combination may modulate a few pathways related to systematic immunity, including ‘Toll-like receptors cascades,’ ‘MAP kinase activation,’ ‘adaptive immune system,’ ‘growth hormone receptor signaling,’ ‘cytokine signaling in immune system,’ and ‘innate immune system’ via Reactome Knowledgebase (Jassal et al., 2020; Figure 3). We believe that all these may account for the therapeutic profiles of the herbal drug combination toward PCM. For instance, ‘Toll-like receptors cascades,’ ‘adaptive immune system,’ ‘cytokine signaling in immune system,’ and ‘innate immune system’ are critical cellular pathways for systematic immunity that are directly associated with the pathogenesis of PCM. Moreover, the ‘MAP kinase activation’ pathway is associated with cellular defense and innate immunity, which are also crucial for the development and inflammatory conditions of PCM.

Figure 2 with 1 supplement see all

Download asset Open asset

Figure 3

Download asset Open asset

Efficacy was assessed every two cycles, and the results were summarized after 6 mo of treatment. The baseline characteristics are shown in Supplementary file 1A. Strikingly, our results demonstrated that a few inflammatory cytokines in the serum, including IL-2, IL-4, IL-6, IFN-γ, IL-1β, and TNF-α, were significantly downregulated in PCM patients after treatment of herbal drug combination compared to the CG treated with methylprednisolone (Figure 4). We chose to measure these cytokines in the experiments because they are often regarded as serum cytokine markers during the pathogenic development of PCM (Liu et al., 2018). In addition, we found that the serum levels of IgA and IgG were markedly suppressed in the treatment group of herbal drug combination compared to the CG (Figure 5). Note that both IgA and IgG have been found to be crucial diagnostic serum markers for PCM patients (Xing et al., 2022). Moreover, IgA is regarded as a major component of mucosal immunity that is closely related to the pathogenesis of PCM (Betts et al., 2018; Bharathan and Mullarky, 2011). Therefore, our data suggests that the herbal drug combination may enable the regulation of mucosal immunity and consequently downregulate IgA and IgG serum level. Furthermore, we conducted the standard Quality of Life questionnaire studies for PCM patients in the clinical experiment. Our results implicated that symptom score, pain score, and global health status of PCM patients are significantly improved after treatment of the herbal drug combination compared to the CG (Figure 6). It is noteworthy that our results demonstrated that the recurrence rate of PCM patients in the treatment group was reduced to 3.75% compared to the recurrence rate of 12.5% in the CG (Table 1). Moreover, the incidence rate of adverse events of PCM patients in the treatment group was reduced to 6.25% compared to the recurrence rate of 11.25% in the CG (Table 1). In addition, we observed that the clinical symptoms of PCM patients in the EG such as swelling, abscess, and fistula were reversed (Figure 7), Supplementary file 3 after treatment of herbal drug combinations. The clinical symptom score in the EG is ~4.68 compared to the clinical symptom score of ~5.98 in the CG (Table 2). These results suggest that the herbal drug combination may achieve better efficacy for the treatment of PCM compared to methylprednisolone.

Figure 4

Download asset Open asset

Figure 4—source data 1

Patients’ raw data of serum cytokines from the control group (without Traditional Chinese Medicine [TCM] treatment) and the treatment group (with TCM treatment) in the clinical trial (ClinicalTrials.gov: NCT05530226).

https://cdn.elifesciences.org/articles/84414/elife-84414-fig4-data1-v3.zip

Download elife-84414-fig4-data1-v3.zip

Figure 5

Download asset Open asset

Figure 5—source data 1

Patients’ raw data of serum IgA and IgG level from the control group (without Traditional Chinese Medicine [TCM] treatment) and the treatment group (with TCM treatment) in the clinical trial (ClinicalTrials.gov: NCT05530226).

https://cdn.elifesciences.org/articles/84414/elife-84414-fig5-data1-v3.zip

Download elife-84414-fig5-data1-v3.zip

Figure 6 with 3 supplements see all

Download asset Open asset

Figure 6—source data 1

Patients’ raw data of symptom scores from the control group (without Traditional Chinese Medicine [TCM] treatment) and the treatment group (with TCM treatment) in the clinical trial (ClinicalTrials.gov: NCT05530226).

https://cdn.elifesciences.org/articles/84414/elife-84414-fig6-data1-v3.zip

Download elife-84414-fig6-data1-v3.zip

Figure 6—source data 2

Patients’ raw data of pain scores from the control group (without Traditional Chinese Medicine [TCM] treatment) and the treatment group (with TCM treatment) in the clinical trial (ClinicalTrials.gov: NCT05530226).

https://cdn.elifesciences.org/articles/84414/elife-84414-fig6-data2-v3.zip

Download elife-84414-fig6-data2-v3.zip

Figure 6—source data 3

Patients’ raw data of quality of life from the control group (without Traditional Chinese Medicine [TCM] treatment) and the treatment group (with TCM treatment) in the clinical trial (ClinicalTrials.gov: NCT05530226).

https://cdn.elifesciences.org/articles/84414/elife-84414-fig6-data3-v3.zip

Download elife-84414-fig6-data3-v3.zip

Figure 7

Download asset Open asset

With the increasing amount of biomedical data, the traditional drug discovery campaign has been revolutionized with the aid of artificial intelligence techniques to accelerate the process and reduce the cost (Yang et al., 2009). In recent years, knowledge graph, a technique that can provide structured relations among entities and unstructured semantic relations associated with entities, has been introduced into the domain of drug discovery (Zeng et al., 2022). The advantage of employing knowledge graph for drug discovery lies in the capabilities of revealing structured associations between drug entities, cellular targets, biological pathways, and phenotypes for human disorders. This is useful for scientists to identify new indications or phenotypes for existing drugs or so-called drug repurposing. With the aid of scoring function or recommendation system, knowledge graph can also be used to design and identify drug combinations for a specific disease. Herein, for the first time, we introduced and employed the concept of knowledge graph to identify herbal drug combinations for the severe PCM with unmet medical needs.

Although the pathogenesis of PCM remains largely unclear, there have been numerous reports implicating that the overactivation of immunoinflammatory pathways plays an important role in the development of PCM (Liu et al., 2015). The major advantage of using TCM is that herbal drug combination can hit multiple discrete targets related to immunoinflammatory pathways with improved efficacy and reduced toxicity. Herein, for the first time, we showcase an example that identifies an herbal drug combination via knowledge graph toward PCM. In contrast to using the conventional principle of ‘syndrome differentiation,’ our knowledge graph consisting of intricate relations between herbal drug entities and immunotherapy targets coupled with scoring functions is able to automatically identify novel herbal drug combinations that can hit most discrete targets, making this strategy unique in the TCM community. This is because PCM is a rather complex disease that pathogenesis may involve multiple targets and immunoinflammatory pathways. Hence, we made the hypothesis that drug combinations that can act on most discrete targets or pathways related to PCM might be more effective to alleviate the symptoms. Although we acknowledge that the inclusion of chemical ingredients from the herbal drugs may impact the outcome of our analysis and design, unfortunately, the inclusion of chemical ingredients in the knowledge graph is rather technically difficult due to the limited and incomplete datasets for the herbal drugs in the field of TCM. Nevertheless, our strategy captures the prominent feature of design for drug combinations toward a complex disease such as PCM. In the future, we plan to include multiple types of omics data such as genomic, transcriptomic, proteomic, metagenomic, or metabolomics data into the knowledge graph to reveal novel targets and enable novel drug discovery.

We want to remind the reader that a third arm (a placebo group) added to the clinical study might be useful to fully reveal the therapeutic effects of the herbal drug combination. Unfortunately, we were unable to add a third arm due to some ethical concerns. This is because PCM is a rather acute, severe, and intense clinical manifestation of breast. Without any treatment, the serious inflammatory condition of PCM may quickly advance into breast cancer. Therefore, this is a limitation of our clinical study, and we hope to design more delicate clinical trials in the future to demonstrate the therapeutic effects of herbal drug combination. Nevertheless, the detailed analysis of three indicators, including symptom score, pain score, and score of life quality, between the EG treated with herbal drug combination and the CG treated with methylprednisolone in the clinical study suggests that the herbal drug combination might be more effective in reversing the clinical conditions of PCM compared to the methylprednisolone treatment (see Figure 6—figure supplement 1, Figure 6—figure supplement 2, Figure 6—figure supplement 3).

In this study, our results revealed that the herbal drug combination identified by knowledge graph could suppress a few key immunoinflammatory cytokines, enhance the systematic immune levels, and significantly reduce the recurrence rates of PCM patients. Note that recurrence has become one major obstacle after surgical resection for PCM treatment in clinical practice. On the other hand, hormone therapy may increase the risk of side effects for PCM patients. Therefore, our approach of herbal drug combination may provide a new avenue for PCM treatment with less recurrence rate and reduced incidence rate of adverse events.

Figure 1–3 are computational and therefore no data have been generated. In addition, Figure 4—source data 1, Figure 5—source data 1, Figure 6—source data 1, Figure 6—source data 2 and Figure 6—source data 3 contain the numerical data used to generate the figures.

1. 1. Betts CB
   2. Pennock ND
   3. Caruso BP
   4. Ruffell B
   5. Borges VF
   6. Schedin P

   (2018) Mucosal immunity in the female murine mammary gland

   Journal of Immunology 201:734–746.

   https://doi.org/10.4049/jimmunol.1800023

   • PubMed
   • Google Scholar
2. 1. Bharathan M
   2. Mullarky IK

   (2011) Targeting mucosal immunity in the battle to develop a mastitis vaccine

   Journal of Mammary Gland Biology and Neoplasia 16:409–419.

   https://doi.org/10.1007/s10911-011-9233-1

   • PubMed
   • Google Scholar
3. 1. Cheng L
   2. Reddy V
   3. Solmos G
   4. Watkins L
   5. Cimbaluk D
   6. Bitterman P
   7. Ghai R
   8. Gattuso P

   (2015) Mastitis, a radiographic, clinical, and histopathologic review

   The Breast Journal 21:403–409.

   https://doi.org/10.1111/tbj.12430

   • PubMed
   • Google Scholar
4. 1. Cutler M

   (1949) Plasma-cell mastitis; report of a case with bilateral involvement

   British Medical Journal 1:94–96.

   https://doi.org/10.1136/bmj.1.4593.94

   • PubMed
   • Google Scholar
5. 1. Fleming L
   2. Agnew S
   3. Peddie N
   4. Crawford M
   5. Dixon D
   6. MacPherson I

   (2022) The impact of medication side effects on adherence and persistence to hormone therapy in breast cancer survivors: a quantitative systematic review

   Breast 64:63–84.

   https://doi.org/10.1016/j.breast.2022.04.010

   • PubMed
   • Google Scholar
6. 1. Hao YF
   2. Jiang JG

   (2015) Origin and evolution of China Pharmacopoeia and its implication for traditional medicines

   Mini Reviews in Medicinal Chemistry 15:595–603.

   https://doi.org/10.2174/1389557515666150415150803

   • PubMed
   • Google Scholar
7. 1. Jassal B
   2. Matthews L
   3. Viteri G
   4. Gong C
   5. Lorente P
   6. Fabregat A
   7. Sidiropoulos K
   8. Cook J
   9. Gillespie M
   10. Haw R
   11. Loney F
   12. May B
   13. Milacic M
   14. Rothfels K
   15. Sevilla C
   16. Shamovsky V
   17. Shorser S
   18. Varusai T
   19. Weiser J
   20. Wu G
   21. Stein L
   22. Hermjakob H
   23. D’Eustachio P

   (2020) The reactome pathway knowledgebase

   Nucleic Acids Research 48:D498–D503.

   https://doi.org/10.1093/nar/gkz1031

   • PubMed
   • Google Scholar
8. 1. Jin W
   2. Stokes JM
   3. Eastman RT
   4. Itkin Z
   5. Zakharov AV
   6. Collins JJ
   7. Jaakkola TS
   8. Barzilay R

   (2021) Deep learning identifies synergistic drug combinations for treating COVID-19

   PNAS 118:e2105070118.

   https://doi.org/10.1073/pnas.2105070118

   • PubMed
   • Google Scholar
9. 1. Li S
   2. Zhang B
   3. Jiang D
   4. Wei Y
   5. Zhang N

   (2010) Herb network construction and co-module analysis for uncovering the combination rule of traditional Chinese herbal formulae

   BMC Bioinformatics 11 Suppl 11:S6.

   https://doi.org/10.1186/1471-2105-11-S11-S6

   • PubMed
   • Google Scholar
10. 1. Liu Y
    2. Zhang J
    3. Zhou YH
    4. Jiang YN
    5. Zhang W
    6. Tang XJ
    7. Ren Y
    8. Han SP
    9. Liu PJ
    10. Xu J
    11. He JJ

    (2015)

    IL-6/stat3 signaling pathway is activated in plasma cell mastitis

    International Journal of Clinical and Experimental Pathology 8:12541–12548.

    • PubMed
    • Google Scholar
11. 1. Liu Y
    2. Zhang J
    3. Zhou Y-H
    4. Zhang H-M
    5. Wang K
    6. Ren Y
    7. Jiang Y-N
    8. Han S-P
    9. He J-J
    10. Tang X-J

    (2018) Activation of the IL-6/JAK2/STAT3 pathway induces plasma cell mastitis in mice

    Cytokine 110:150–158.

    https://doi.org/10.1016/j.cyto.2018.05.002

    • PubMed
    • Google Scholar
12. 1. Liu Y
    2. Sun Y
    3. Zhou Y
    4. Tang X
    5. Wang K
    6. Ren Y
    7. He J

    (2020) Sinomenine hydrochloride inhibits the progression of plasma cell mastitis by regulating IL-6/JAK2/STAT3 pathway

    International Immunopharmacology 81:106025.

    https://doi.org/10.1016/j.intimp.2019.106025

    • PubMed
    • Google Scholar
13. 1. Nan Q
    2. Qinnan W
    3. Chaoqun MA
    4. Dexuan C
    5. Haidong C
    6. Yaoyao LU

    (2022) Mechanism underlying efficacy of shugan sanjie decoction on plasma cell mastitis, based on network pharmacology and experimental verification

    Journal of Traditional Chinese Medicine = Chung i Tsa Chih Ying Wen Pan 42:400–407.

    https://doi.org/10.19852/j.cnki.jtcm.20220311.004

    • PubMed
    • Google Scholar
14. 1. Reimand J
    2. Isserlin R
    3. Voisin V
    4. Kucera M
    5. Tannus-Lopes C
    6. Rostamianfar A
    7. Wadi L
    8. Meyer M
    9. Wong J
    10. Xu C
    11. Merico D
    12. Bader GD

    (2019) Pathway enrichment analysis and visualization of omics data using G: profiler, GSEA, cytoscape and enrichmentmap

    Nature Protocols 14:482–517.

    https://doi.org/10.1038/s41596-018-0103-9

    • PubMed
    • Google Scholar
15. 1. Tu Y

    (2011) The discovery of artemisinin (qinghaosu) and gifts from Chinese medicine

    Nature Medicine 17:1217–1220.

    https://doi.org/10.1038/nm.2471

    • PubMed
    • Google Scholar
16. 1. Tu Y

    (2016) Artemisinin-A gift from traditional chinese medicine to the world (nobel lecture)

    Angewandte Chemie 55:10210–10226.

    https://doi.org/10.1002/anie.201601967

    • PubMed
    • Google Scholar
17. 1. Xing M
    2. Zhang S
    3. Zha X
    4. Zhang J

    (2022) Current understanding and management of plasma cell mastitis: can we benefit from what we know?

    Breast Care 17:321–329.

    https://doi.org/10.1159/000517572

    • PubMed
    • Google Scholar
18. 1. Yang Y
    2. Adelstein SJ
    3. Kassis AI

    (2009) Target discovery from data mining approaches

    Drug Discovery Today 14:147–154.

    https://doi.org/10.1016/j.drudis.2008.12.005

    • PubMed
    • Google Scholar
19. 1. Ye Q
    2. Hsieh CY
    3. Yang Z
    4. Kang Y
    5. Chen J
    6. Cao D
    7. He S
    8. Hou T

    (2021) A unified drug-target interaction prediction framework based on knowledge graph and recommendation system

    Nature Communications 12:6775.

    https://doi.org/10.1038/s41467-021-27137-3

    • PubMed
    • Google Scholar
20. 1. Yin X
    2. Li W
    3. Liang T
    4. Lu B
    5. Yue H
    6. Li S
    7. Zhong VW
    8. Zhang W
    9. Li X
    10. Zhou S
    11. Mi Y
    12. Wu H
    13. Xu S

    (2022) Effect of electroacupuncture on insomnia in patients with depression: a randomized clinical trial

    JAMA Network Open 5:e2220563.

    https://doi.org/10.1001/jamanetworkopen.2022.20563

    • PubMed
    • Google Scholar
21. 1. Yu J
    2. Bao S
    3. Yu S
    4. Zhang D-Q
    5. Loo WTY
    6. Chow LWC
    7. Su L
    8. Cui Z
    9. Chen K
    10. Ma L-Q
    11. Zhang N
    12. Yu H
    13. Yang Y-Z
    14. Dong Y
    15. Yip AYS
    16. Ng ELY

    (2012) Mouse model of plasma cell mastitis

    Journal of Translational Medicine 10 Suppl 1:S11.

    https://doi.org/10.1186/1479-5876-10-S1-S11

    • PubMed
    • Google Scholar
22. 1. Zeng X
    2. Tu X
    3. Liu Y
    4. Fu X
    5. Su Y

    (2022) Toward better drug discovery with knowledge graph

    Current Opinion in Structural Biology 72:114–126.

    https://doi.org/10.1016/j.sbi.2021.09.003

    • PubMed
    • Google Scholar
23. 1. Zhang J
    2. Xu J
    3. Zhang J
    4. Ren Y

    (2020a) Chinese herbal compound combined with western medicine therapy in the treatment of plasma cell mastitis: a protocol for systematic review and meta-analysis

    Medicine 99:e22858.

    https://doi.org/10.1097/MD.0000000000022858

    • PubMed
    • Google Scholar
24. 1. Zhang Y
    2. Yao Y
    3. Chen P
    4. Liu Y
    5. Zhang H
    6. Liu H
    7. Liu Y
    8. Xu H
    9. Tian X
    10. Wang Z
    11. Chu P
    12. Zhao D
    13. Liu H
    14. Zhang C
    15. Chen S
    16. Zhao Y
    17. Liu C
    18. Yang Y

    (2020b) Checkpoint therapeutic target database (CKTTD): the first comprehensive database for checkpoint targets and their modulators in cancer immunotherapy

    Journal for Immunotherapy of Cancer 8:e001247.

    https://doi.org/10.1136/jitc-2020-001247

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "A herbal drug combination identified by knowledge graph alleviates the clinical symptoms of plasma cell mastitis patients: a non-randomized controlled trial" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Mone Zaidi as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

(1) Please address Reviewer 1's comments about the patient groups and confounding factors as well as expand the rationale for the evaluation of the clinical trial as noted for example, by reviewer 2's comments about the combination drug score.

(2) Please address the relatively minor comments of all three reviewers.

Reviewer #1 (Recommendations for the authors):

Figure 2 looks like it could be moved to supplementary data.

Figure 3 text was too small to read.

The presented changes in figures 5 and 6, while statistically significant do not demonstrate significant difference. Perhaps a third, placebo, arm is needed whereby the steroids and the herbal groups induce immune changes not seen in the placebo group? That data would be clear and convincing, but as presented it is difficult to understand if this is a medication effect o=r the natural course of disease in these patients.

Reviewer #2 (Recommendations for the authors):

Below are some specific points for the manuscript,

1. It would be necessary to elaborate the advantages of employing knowledge graph to design drug combinations for a specific human disease in the Discussion section of the manuscript.

2. In Figure 4, the authors claimed that 'a few pathways related to systematic immunity' were identified. In page 8, line 158, the authors should discuss in length for the specific roles of these pathways in PCM pathogenesis.

3. The drug combination in the clinical trial didn't obtain the highest score (Figure 2). The authors need to explain this in the Discussion section.

4. Figure 1, Supplement 1, the authors need to explain the color scheme (pink, yellow and green etc) for the entities.

5. Page 6, line 133, please explain and specify the term 'syndrome differentiation'. The reader who are not in TCM domain may not understand the term.

6. In page 6, the authors claimed that the scoring system is able to identify 'herbal drug combinations that are able to hit most discrete cellular targets'. The authors need to explain and elaborate why this criteria is crucial for the identification of drug combinations for PCM in the Discussion section.

7. The authors need to provide more details for their clinical study. For example, more details are required for the patients in Figure 6.

Reviewer #3 (Recommendations for the authors):

This is a rather interesting study that introduced a new concept in the domain of TCM and herbal drugs. Below are some specific suggestions which I think may help to improve the manuscript,

1. The herbal drug combination used in the clinical study didn't obtain the highest score during the scoring process (Figure 2). The authors need to explain and elaborate this in a paragraph.

2. In Figure 3, the network diagram analysis from the knowledge graph revealed that NLPR3, IL-17, TLR4, STAT3, IL-6, iNOS and TLR2 are possible cellular targets for the herbal drug combination. The authors need to elaborate or provide some references that NLPR3, IL-17, TLR4, STAT3, IL-6, iNOS and TLR2 are relevant to PCM pathogenesis.

3. The authors need to explain and elaborate 'penalty function', in page 14.

4. Line 326, the authors mean 'validated via cellular model'?

5. Line 329, the author refers to 'systematic immunity'?

6. Page 17, line 347, 'The herbal drug combination was prepared as granules in the following formulae', The authors need to explain how this formula was determined.

7. In Figure 8, the authors need to provide more details for the clinical pictures in a paragraph. For example, age, treatment period and brief descriptions of the symptoms etc.

8. The authors need to explain why the drug combination consisting of eight herbal drugs is selected in their study. Page 7, line 137, the authors made a statement that 'eight drugs are regarded as essence combination'. The authors need to explain and elaborate this point in the manuscript.

9. Is it feasible that the authors could include chemical ingredients of the herbal drugs into their knowledge graph for the identification of herbal drug combination? The authors should elaborate more for this point in the manuscript.

Reviewer #1 (Recommendations for the authors):

Figure 2 looks like it could be moved to supplementary data.

This has been fixed. Per the referee’s suggestion, we have move Figure 2 to supplementary data.

Figure 3 text was too small to read.

This has been fixed. Per the referee’s comment, we have rearranged and enlarged the text in Figure 3 (now Figure 2 in draft-track-change). Please see draft-track-change for more details.

The presented changes in figures 5 and 6, while statistically significant do not demonstrate significant difference. Perhaps a third, placebo, arm is needed whereby the steroids and the herbal groups induce immune changes not seen in the placebo group? That data would be clear and convincing, but as presented it is difficult to understand if this is a medication effect o=r the natural course of disease in these patients.

We want to thank the referee for this very insightful and expert comment. We fully agree with the referee that a third arm (placebo group) added in the clinical trial may be useful to further reveal the therapeutic effects of herbal drug combinations. Indeed, we met some difficulties to add the placebo group in the clinical trial and unfortunately, we were unable to obtain the ethnical approval from the Research Ethnical Committee of our university should we want to add the placebo group in the trial. This is because plasma cell mastitis (PCM) is a rather acute, severe and intense clinical manifestation of breast. Without any treatment, the serious inflammatory condition of PCM may quickly advance into breast cancer. In addition, the patients may feel tremendous pain without treatment due to the quick development of PCM. Therefore, the Research Ethnical Committee raised concerns about the placebo group and we were unable to add the placebo group in the trial. Nevertheless, we fully agree with the referee that a third arm could be useful and we elaborate this limitation of our clinical study in a paragraph in the Discussion section of our manuscript. Moreover, we conducted additional analysis of the clinical data. The detailed analysis of three indicators including symptom score, pain score and score of life quality between the experiment group treated with herbal drug combination and the control group treated with methylprednisolone in the clinical study suggest that the herbal drug combination might be more effective to reverse the clinical conditions of PCM as compared to the methylprednisolone treatment (see Figure 6 —figure supplement 1-2-3). Please see below or draft-track-change, line 234-246 for more details.

“We want to remind the reader that a third arm (a placebo group) added in the clinical study might be useful to fully reveal the therapeutic effects of the herbal drug combination. Unfortunately, we were unable to add a third arm due to some ethnical concerns. This is because plasma cell mastitis (PCM) is a rather acute, severe and intense clinical manifestation of breast. Without any treatment, the serious inflammatory condition of PCM may quickly advance into breast cancer. Therefore, this is a limitation of our clinical study and we hope to design more delicate clinical trial in the future to demonstrate the therapeutic effects of herbal drug combination. Nevertheless, the detailed analysis of three indicators including symptom score, pain score and score of life quality between the experiment group treated with herbal drug combination and the control group treated with methylprednisolone in the clinical study suggest that the herbal drug combination might be more effective to reverse the clinical conditions of PCM as compared to the methylprednisolone treatment (see Figure 6 —figure supplement 1-2-3).”

Reviewer #2 (Recommendations for the authors):

Below are some specific points for the manuscript,

1. It would be necessary to elaborate the advantages of employing knowledge graph to design drug combinations for a specific human disease in the Discussion section of the manuscript.

Many thanks for this critical and helpful comment. Per the referee’s suggestion, we have elaborated the advantage of the knowledge graph towards the design of drug combinations in the Discussion section of the manuscript. Please see below or draft-track-change, line 214-222 for more details.

“The advantage of employing knowledge graph for drug discovery lies in the capabilities of revealing structured associations between drug entities, cellular targets, biological pathways and phenotypes for human disorders. This is useful for scientists to identify new indications or phenotypes for existing drugs, or so-called drug repurposing. With the aid of scoring function or recommendation system, knowledge graph can also be used to design and identify drug combinations for a specific disease. Herein, for the first time, we introduced and employed the concept of knowledge graph to identify herbal drug combinations for the severe Plasma cell mastitis (PCM) with unmet medical needs.”

2. In Figure 4, the authors claimed that 'a few pathways related to systematic immunity' were identified. In page 8, line 158, the authors should discuss in length for the specific roles of these pathways in PCM pathogenesis.

We want to thank the referee for this insightful comment. This has been fixed. Per the referee’s suggestion, we have discussed in length for the specific roles of the related pathways in PCM. Please see below or draft-track-change, line 214-222 for more details.

“For instance, ‘Toll-like receptors cascades’, ‘Adaptive immune system’, ‘Cytokine signaling in immune system’ and ‘Innate immune system’ are critical cellular pathways for systematic immunity which are directly associated with the pathogenesis of PCM. Moreover, the ‘MAP kinase activation’ pathway is associated with cellular defense and innate immunity which are also crucial for the development and inflammatory conditions of PCM.”

3. The drug combination in the clinical trial didn't obtain the highest score (Figure 2). The authors need to explain this in the Discussion section.

Thanks for this expert and helpful comment. This has been fixed. The drug combination was chosen for the further clinical study for two reasons. First, this drug combination was among top twenty combinations in each round of calculations. Second, we asked experts in TCM to inspect the top twenty combinations on the basis of ‘syndrome differentiation’ as described in Pharmacopoeia of China and finally the combination consisting of eight herbal drug entities including ‘Fructus forsythiae’, ‘Herba violae’, ‘Uniflower swisscentaury root’, ‘Danshen’, ‘Astragalus’, ‘Taraxacum’, ‘Liquorice’ and ‘Honeysuckle’ was selected for further clinical study. Please see below or draft-track-change, line 144-151 for more details.

“The drug combination was chosen for the further clinical study for two reasons. First, this drug combination was among top twenty combinations in each round of our calculations. Second, we asked experts in TCM to inspect the top twenty combinations on the basis of ‘syndrome differentiation’ as described in Pharmacopoeia of China and finally the combination consisting of eight herbal drug entities including ‘Fructus forsythiae’, ‘Herba violae’, ‘Uniflower swisscentaury root’, ‘Danshen’, ‘Astragalus’, ‘Taraxacum’, ‘Liquorice’ and ‘Honeysuckle’ was selected for further clinical study (see Figure 2).”

4. Figure 1, Supplement 1, the authors need to explain the color scheme (pink, yellow and green etc) for the entities.

Thank you for this critical comment. This has been fixed. We have explained the various color scheme in the figure legend. Please see page 22, in draft-track-change for more details.

5. Page 6, line 133, please explain and specify the term 'syndrome differentiation'. The reader who are not in TCM domain may not understand the term.

We are grateful for this critical comment. This has been fixed. Per the referee’s suggestion, we have explained the term ‘syndrome differentiation’ in the manuscript. Please see Please see below or line 134-138, in draft-track-change for more details.

“Here, syndrome differentiation refers to the very basic principle of identifying and treating disease in TCM. Syndrome (Zheng) is the presentation of the pathological changes during a specific disease course including the location, cause and nature of a disease. Moreover, ‘Jun-Chen-Zuo-Shi’ refers to the rules guided by syndrome differentiation to select multiple herbal drug entities to treat a specific disease in TCM.”

6. In page 6, the authors claimed that the scoring system is able to identify 'herbal drug combinations that are able to hit most discrete cellular targets'. The authors need to explain and elaborate why this criteria is crucial for the identification of drug combinations for PCM in the Discussion section.

Many thanks for this critical comment. Per the referee’s suggestion, we have explained why the criterion of the scoring system is crucial for the identification of drug combinations in the manuscript. Please see below or line 249-252 in draft-track-change for more details.

“This is because PCM is a rather complex disease which pathogenesis may involve multiple targets and immunoinflammatory pathways. Hence, we made the hypothesis that drug combinations that can act on most discrete targets or pathways related to PCM might be more effective.”

7. The authors need to provide more details for their clinical study. For example, more details are required for the patients in Figure 6.

This has been fixed. We have provided detailed information for the six patients in Figure 6 as supplemental materials (SI_patients_information).

Reviewer #3 (Recommendations for the authors):

This is a rather interesting study that introduced a new concept in the domain of TCM and herbal drugs. Below are some specific suggestions which I think may help to improve the manuscript,

1. The herbal drug combination used in the clinical study didn't obtain the highest score during the scoring process (Figure 2). The authors need to explain and elaborate this in a paragraph.

Thanks for this expert and helpful comment. This has been fixed. The drug combination was chosen for the further clinical study for two reasons. First, this drug combination was among top twenty combinations in each round of calculations. Second, we asked experts in TCM to inspect the top twenty combinations on the basis of ‘syndrome differentiation’ as described in Pharmacopoeia of China and finally the combination consisting of eight herbal drug entities including ‘Fructus forsythiae’, ‘Herba violae’, ‘Uniflower swisscentaury root’, ‘Danshen’, ‘Astragalus’, ‘Taraxacum’, ‘Liquorice’ and ‘Honeysuckle’ was selected for further clinical study. Please see below or draft-track-change, line 144-151 for more details.

“The drug combination was chosen for the further clinical study for two reasons. First, this drug combination was among top twenty combinations in each round of our calculations. Second, we asked experts in TCM to inspect the top twenty combinations on the basis of ‘syndrome differentiation’ as described in Pharmacopoeia of China and finally the combination consisting of eight herbal drug entities including ‘Fructus forsythiae’, ‘Herba violae’, ‘Uniflower swisscentaury root’, ‘Danshen’, ‘Astragalus’, ‘Taraxacum’, ‘Liquorice’ and ‘Honeysuckle’ was selected for further clinical study (see Figure 2).”

2. In Figure 3, the network diagram analysis from the knowledge graph revealed that NLPR3, IL-17, TLR4, STAT3, IL-6, iNOS and TLR2 are possible cellular targets for the herbal drug combination. The authors need to elaborate or provide some references that NLPR3, IL-17, TLR4, STAT3, IL-6, iNOS and TLR2 are relevant to PCM pathogenesis.

We are grateful for this critical and helpful comment. This has been fixed. We have provided references for these putative targets. Please see page 8, line 161-162 in draft-track-change for more details.

3. The authors need to explain and elaborate 'penalty function', in page 14.

This has been fixed. We have explained the ‘penalty function’ in the manuscript. Please see below or page 13, line 339-342 in draft-track-change for more details.

“Herein, the penalty function is used to ensure that the herbal drug entities in the combination didn’t violate the contraindication rules in Pharmacopoeia of China according to the medicinal attributes of herbal drugs.”

4. Line 326, the authors mean 'validated via cellular model'?

This has been fixed. Here we mean ‘validated via cellular model’.

5. Line 329, the author refers to 'systematic immunity'?

This has been fixed. Herein we refer to 'systematic immunity'.

6. Page 17, line 347, 'The herbal drug combination was prepared as granules in the following formulae', The authors need to explain how this formula was determined.

This has been explained. Please see below or page 19, line 393-398 in draft-track-change for more details.

“The formula was determined by TCM experts on the basis of ‘syndrome differentiation’ as described in Pharmacopoeia of China. Furthermore, the herbal drug combination in the form of granules was provided and prepared by Shengjing Hospital Affiliated to China Medical University according to the standard requirement of clinical study by National Medical Products Administration (NMPA).”

7. In Figure 8, the authors need to provide more details for the clinical pictures in a paragraph. For example, age, treatment period and brief descriptions of the symptoms etc.

This has been fixed. We have provided detailed information for the six patients in Figure 6 as supplemental materials (SI_patients_information).

8. The authors need to explain why the drug combination consisting of eight herbal drugs is selected in their study. Page 7, line 137, the authors made a statement that 'eight drugs are regarded as essence combination'. The authors need to explain and elaborate this point in the manuscript.

We are thankful for this critical and helpful comment. This has been fixed. We chose to identify drug combinations with eight entities because ‘formulae’ consisting of eight drugs are regarded as ‘essence combination’ in TCM field. Please see page 7, line 140-142 in draft-track-change for more details.

9. Is it feasible that the authors could include chemical ingredients of the herbal drugs into their knowledge graph for the identification of herbal drug combination? The authors should elaborate more for this point in the manuscript.

Many thanks for this insightful and helpful comment. We fully agree with the referee that the inclusion of chemical ingredients from the herbal drugs may impact the outcome of our analysis and design. Unfortunately, the inclusion of chemical ingredients in the knowledge graph is rather technically difficult due to the limited and incomplete datasets for the herbal drugs in the field of TCM. Nevertheless, our strategy captures the prominent feature of design for drug combinations towards a complex disease such as PCM. In the future, we plan to include multiple types of omics data such as genomic, transcriptomic, proteomic, metagenomic or metabolomics data into the knowledge graph to reveal novel targets and enable novel drug discovery. Please see page 12, line 252-260 in draft-track-change for more details. We hope this could clarify the concerns.

Author details

1. Caigang Liu

   Cancer Stem Cell and Translation Medicine Lab, Innovative Cancer Drug Research and Development Engineering Center of Liaoning Province, Department of Oncology, Shengjing Hospital of China Medical University, Shenyang, China

   Contribution

   Conceptualization, Resources, Supervision, Funding acquisition, Project administration

   For correspondence

   angel-s205@163.com

   Competing interests

   Senior editor, eLife

   "This ORCID iD identifies the author of this article:" 0000-0003-2083-235X

2. Hong Yu

   Cancer Stem Cell and Translation Medicine Lab, Innovative Cancer Drug Research and Development Engineering Center of Liaoning Province, Department of Oncology, Shengjing Hospital of China Medical University, Shenyang, China

   Contribution

   Data curation, Supervision, Investigation, Project administration

   Contributed equally with

   Guanglei Chen

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-4557-1558

3. Guanglei Chen

   Cancer Stem Cell and Translation Medicine Lab, Innovative Cancer Drug Research and Development Engineering Center of Liaoning Province, Department of Oncology, Shengjing Hospital of China Medical University, Shenyang, China

   Contribution

   Validation, Visualization, Project administration

   Contributed equally with

   Hong Yu

   Competing interests

   No competing interests declared

4. Qichao Yang

   School of Bioengineering, Dalian University of Technology, Dalian, China

   Contribution

   Data curation, Software, Visualization, Methodology

   Competing interests

   No competing interests declared

5. Zichu Wang

   School of Bioengineering, Dalian University of Technology, Dalian, China

   Contribution

   Software, Methodology

   Competing interests

   No competing interests declared

6. Nan Niu

   Cancer Stem Cell and Translation Medicine Lab, Innovative Cancer Drug Research and Development Engineering Center of Liaoning Province, Department of Oncology, Shengjing Hospital of China Medical University, Shenyang, China

   Contribution

   Validation, Project administration

   Competing interests

   No competing interests declared

7. Ling Han

   National Engineering Research Center of Pharmaceutics of Traditional Chinese Medicine, China Resources Sanjiu Medical & Pharmaceutical Co., Ltd, Shenzhen, China

   Contribution

   Resources, Validation, Visualization

   Competing interests

   is an employee of China Resources Sanjiu Medical & Pharmaceutical

8. Dongyu Zhao

   International Cancer Institute, Peking University Health Science Center, Peking University, Beijing, China

   Contribution

   Validation, Investigation, Visualization

   Competing interests

   No competing interests declared

9. Manji Wang

   Shanghai BeautMed Corporation, Shanghai, China

   Contribution

   Validation, Investigation, Visualization

   For correspondence

   147304368@qq.com

   Competing interests

   is an employee of Shanghai BeautMed Corporation

10. Yuanyuan Liu

    1. School of Bioengineering, Dalian University of Technology, Dalian, China
    2. Department of Biology, University of Copenhagen, Copenhagen, Denmark

    Contribution

    Data curation, Software

    Competing interests

    No competing interests declared

11. Yongliang Yang

    School of Bioengineering, Dalian University of Technology, Dalian, China

    Contribution

    Conceptualization, Resources, Supervision, Funding acquisition, Investigation, Methodology, Writing - original draft, Project administration, Writing - review and editing

    For correspondence

    everbright99@foxmail.com

    Competing interests

    Reviewing editor, eLife

    "This ORCID iD identifies the author of this article:" 0000-0003-0449-0599

• 391

  Page views

• 87

  Downloads

• 0

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.