Article pubs.acs.org/jcim
Toward Understanding the Cold, Hot, and Neutral Nature of Chinese Medicines Using in Silico Mode-of-Action Analysis Xianjun Fu,†,‡ Lewis H. Mervin,‡ Xuebo Li,† Huayun Yu,§ Jiaoyang Li,† Siti Zuraidah Mohamad Zobir,‡ Azedine Zoufir,‡ Yang Zhou,† Yongmei Song,† Zhenguo Wang,*,† and Andreas Bender*,‡ †
School of Information Management, Shandong University of Traditional Chinese Medicine, 250355 Jinan, China Centre for Molecular Science Informatics, Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, United Kingdom § College of TCM, Shandong University of Traditional Chinese Medicine, 250355 Jinan, China ‡
S Supporting Information *
ABSTRACT: One important, however, poorly understood, concept of Traditional Chinese Medicine (TCM) is that of hot, cold, and neutral nature of its bioactive principles. To advance the field, in this study, we analyzed compound−nature pairs from TCM on a large scale (>23 000 structures) via chemical space visualizations to understand its physicochemical domain and in silico target prediction to understand differences related to their modes-of-action (MoA) against proteins. We found that overall TCM natures spread into different subclusters with specific molecular patterns, as opposed to forming coherent global groups. Compounds associated with cold nature had a lower clogP and contain more aliphatic rings than the other groups and were found to control detoxification, heat-clearing, heart development processes, and have sedative function, associated with “Mental and behavioural disorders” diseases. While compounds associated with hot nature were on average of lower molecular weight, have more aromatic ring systems than other groups, frequently seemed to control body temperature, have cardio-protection function, improve fertility and sexual function, and represent excitatory or activating effects, associated with “endocrine, nutritional and metabolic diseases” and “diseases of the circulatory system”. Compounds associated with neutral nature had a higher polar surface area and contain more cyclohexene moieties than other groups and seem to be related to memory function, suggesting that their nature may be a useful guide for their utility in neural degenerative diseases. We were hence able to elucidate the difference between different nature classes in TCM on the molecular level, and on a large data set, for the first time, thereby helping a better understanding of TCM nature theory and bridging the gap between traditional medicine and our current understanding of the human body.
■
INTRODUCTION
of clinic experience, the nature theory was developed to be a very important property of the Chinese medicines (CMs) and became the basis of standard therapeutic guidelines, which goes beyond immediately perceivable sensations following administration.7 The therapeutic effect of CMs depends mainly on the nature of the drugs as well as the processes they regulate to recover the balance between Yin and Yang in the human body.8,9 According to TCM, the rationale for the correct remedy selection is based upon a corresponding TCM syndrome (Zheng, , or pattern).7 A patient will present with a syndrome upon disruption of Yin-yang balance, which may be caused either by external or internal pathological factors. This can be regarded as clinical phenotype, such as Cold or Hot Syndrome.2,7 The standard therapeutic guideline used to treat cold or hot syndrome is to “cure cold syndrome by medication with hot
Traditional Chinese Medicine (TCM) has been used to treat and prevent diseases for thousands of years and is based on a unique system of theory, diagnosis, and treatment (illustrated in Figure 1).1 Most attention has so far been paid to the theories related to the understanding of the human body, such as Yin-yang ( ) ).1,2 Secondary theories related and five elements (Wuxing, to the medicines, such as nature (also called Property, Yaoxing, , and Siqi, ), Five Tastes (Wuwei, ), and Meridian ) have been largely ignored for a long Tropism (Guijing, time.3,4 The theory of nature, which can be classified into three types, cold, hot, and neutral, are believed to mainly originate from the reactions of the body after the administration of a specific medicine.5,6 For example, chewing a mint (Mentha spicata) leaf elicits a cold feeling, while masticating a piece of ginger (Zingiber off icinale) root leads to a hot sensation, and a neutral remedy, such as Wolfberry (Lycium barbarum), may not be associated with an obvious cold or hot feeling.6 However, with a long history © 2017 American Chemical Society
Received: November 30, 2016 Published: March 3, 2017 468
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling
Figure 1. Interactions between the theory of body dimension and medicines theory. The theory system of body dimension includes the Yin Yang system and five elements (Wuxing). The medicine theory system includes nature theory (Siqi), Five tastes theory (Wuwei), and others (not presented here). Five tastes are corresponding to five elements, and for example, medicines with sweet taste are considered to treat symptoms related to spleen (earth). Nature theory is related to Yin Yang theory, and when Yang is stronger than Yin, then patients present hot symptoms, and according to this theory medicines with cold nature would be used to treat it. On the other hand, if Yin is stronger than Yang, then patients present cold symptoms, and medicines with hot nature would be used to treat it. Medicines with neutral nature can be chosen to complement hot or cold medicines to treat different symptoms.
Figure 2. Overview of the workflow of this study. 1. Collecting compound information on Chinese Medicines (CMs). 2. Collecting nature information on Chinese Medicines and grouping compounds into cold, hot, and neutral types. 3. Screening compounds by analysis workflow (see details in Figure S1). 4. Analysis of chemical space. 5. Analysis of bioactivity through chemogenomic methods. 6. Association relationship analysis between molecular structure of orphan compounds taken from TCM databases, predicted targets, and traditional cold, hot, and neutral nature.
nature” and to “cure hot syndrome by medication with cold nature”.7 This therapeutic practice has been validated and developed over thousands of years, and most CMs have thus been labeled with different nature types as a result of this repeated clinical practice.5 The area of understanding the hot, cold, and neutral nature of TCM has attracted some attention recently. Mathematical models, such as the three-element model, were employed to explain the different biomarker levels, that is, body-condition elements, like triiodothyronine, thyroid stimulating hormone, etc., between hot and cold nature medicines.10 Recent literature
suggests that the discrimination mechanism of cold and hot nature is closely related to energy metabolism and may be objectively represented by the temperature tropism of animals.11,12 In particular, the changes of energy metabolism including ATPase activity and oxygen consumption may be plausibly affected by the cold and hot nature differences of Mahuang and Maxingshigan decoctions.13 Traditional Chinese medicines (TCMs) with cold nature, namely, Rhei Radix et Rhizoma and Coptidis rhizoma, can reduce the temperature of fever rats induced by yeast, an effect which may be related to the regulation of transient receptor potential ion channel proteins 469
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling (TRPs) in the hypothalamus and dorsal root ganglia.14 Other authors have suggested that inflammation and the immune system regulation are closely affected by hot natured CMs, while cold CMs have the potential to impact cell growth, proliferation and development.6 Self-organizing maps (SOM) have been used for the prediction of hot and cold natures of CMs from their constituent compounds and for the classification of CM mixtures.15 Literature also proposed that generally neutralnatured CMs present characteristic action of “two-way application and conditioned dominance”, promoting blood circulation and removing blood stasis in cold and hot blood stasis syndrome.16 However, despite the above advances, little is known about the differences of hot, neutral, and cold TCMs on both the chemical, as well as the mechanistic (i.e., protein target) level, which is, hence, the purpose of the current study. Given that we attempted to understand the nature of those classes on a general level, we did not attempt to perform experimental studies (which would have been impossible to perform on the scale of thousands of formulations examined here) but rather used computational methods, such as analyses of chemical space, property prediction, and in silico target prediction and mode-of-action (MoA) analysis instead.17 Given that chemical structure is needed for in silico MoA, a database comprising as much information in this domain as possible is a necessary first step,5 and the subsequent analysis also needs to be tailored to this particular area (the workflow performed here is visualized in Figure 2). The overall nature of an herb is a result of all the compounds it contains. Given that the composition in each herb is rather complex (and usually not even known in detail), it is very difficult (or even impossible) to identify the exact compounds, which are responsible for its overall nature. Hence, in the present study, we associated all compounds present in a given herb with its nature. While this indeed represents an oversimplification, given our lack of more detailed knowledge would otherwise prevent any analysis of hot, cold, and neutral Traditional Chinese Medicine on the molecular level, of the type we have performed here. In this work, we have compiled a database comprising, in total, more than 23 000 compounds annotated with their nature, which we have used for further analysis and which, to the knowledge of the authors, is the largest database of this kind. Since chemical structure is the molecular basis for the activity of compounds,18 characterizing the molecular structure is of central importance for understanding the nature of CMs further.19 To this end, we have in a first step analyzed physicochemical properties, Murcko scaffolds, and visualized CMs in chemical space to identify patterns among the different nature groups.19,20 This includes visualizations to understand nonlinear structure−nature correlations21 and data-mining methods, such as association rule mining22 to understand associations between molecular structure, targets and pathways (for the latter two see below) and the cold, hot and neutral nature of CMs. To understand the proteins which CMs from different nature classes modulate better, we have performed in silico target prediction. In silico protein target prediction is a well-established computational technique which is able to predict protein targets for orphan compounds, based on ligand structure.23−26 This approach offers an avenue to infer target-ligand associations by utilizing known bioactivity information,21 which is now available for millions of compounds in databases, such as ChEMBL or PubChem.21 Various methods of in silico target prediction have been used in TCM before,2,27 based on increasingly more
available information about the chemical constitution of ingredients in electronic form. In this study, a previously benchmarked target prediction protocol has been employed for the prediction of a range of targets to annotate the traditional cold, hot, and neutral nature of CM on the MoA level.23
■
METHODS Compound Data and Analysis Workflow. TCM compounds were obtained from TCMD28 (version 2009) in SD format. An analysis workflow with multimethods was built to screen the compounds responsible for the effect of nature. First, ChemAxon Structure Checker29 was employed to preprocess the compounds include “Wedge Error”, “Ring Strain Error”, “Valence Error”, and “Ring Strain Error” set to check and fix. Next, compounds annotated with their related nature were extracted from the herbs according to the “Chinese Pharmacopoeia”30 and “Chinese Materia Medica”.31 Since exhaustive experimental testing of intestinal absorption for the entire compound set is a time-consuming task, compounds with poor intestinal absorption were discarded using a passive intestinal absorption (PIA) model in Discovery studio 4.5 on a per compound basis.32 To remove redundancy in the data sets (i.e., large number of compounds with highly similar structures), in the next step, DataWarrior21 was used to cluster compounds based on their fingerprint similarities using the SkelSphere descriptor, where the cluster size was set to 1500. All cluster members of clusters where more than 80% of its members were annotated with the same nature type were selected. A total of 23 033 compounds from 2444 CMs were obtained. The analysis workflow screened compounds with several methods (Figure S1). First, preprocessing of the compounds, including the structure checking and fixing by ChemAxon resulted in 22 073 compounds. Then, the annotations of 1908 multinature compounds and 6798 nature-free compounds were discarded leaving 13 367 compounds in the analysis. Next, 7992 compounds with good to moderate levels of absorption were selected. Finally, compounds clustered on their calculated similarities in DataWarrior resulted in 2012 compounds in total, comprising 1033 compounds associated with cold nature, 763 compounds associated with hot nature, and 216 compounds associated with neutral nature, which were subjected to further studies. Hence, the compounds with similar chemical structures associated with different nature were also discarded and the screened compounds show low degree of overlap of chemical space. Chemical Space Visualizations and Similarity Measures. Chemical space visualizations of 2012 compounds was produced by DataWarrior21 using the 2D-Rubber Band Scaling (2D-RBS) method. The SkelSpheres descriptor was selected as the compound fingerprint used and marker colors and shapes represent the different nature types. Molecular Descriptors Calculation and Analysis. DataWarrior was used to calculate 26 molecular descriptors (Total Molweight, Molweight, Absolute Weight, clogP, clogS, HAcceptors, H-Donors, Total Surface Area, Polar Surface Area, Druglikeness, Shape Index, Molecular Flexibility, Molecular Complexity, Non-H Atoms, Non-C/H Atoms, Electronegative Atoms, Stereo Centers, Rotatable Bonds, Rings, Aromatic Rings, Aromatic Atoms, sp3-Atoms, Symmetric atoms Amides Amines, Alkyl-Amines, Aromatic Nitrogens, Basic Nitrogens, Acidic Oxygens).21 Principal component analysis (PCA)33 was performed to identify patterns of the descriptors in the three different natures and conducted with RStudio (version 470
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling
Figure 3. Chemical space visualizations of 2012 compounds associated with hot, cold and neutral herbs from TCM. (a) Represents the view depicting the chemical space of all molecules. Similar neighbors are connected with a connecting line and the markers that represent the molecules are colored by different nature, and highlighting that compounds with similar nature type are clustered together. (b−d) Tree views that show the direct neighbors of the chosen current molecule in plot a with different nature, showing both higher chemical similarities while being associated with the same nature group. This figure reveals that compounds with same nature type are clustered together and implied higher chemical similarities in the same nature group. However, clusters are highly local, indicating that overall similarities of compounds associated with the same nature class are rather low and only local similarities are present.
0.99.473)34 using the prcomp function and ggbiplot packages.35 A biplot was adopted to represent the compounds on a twodimensional space derived from principal component analyses (PCA) based on the aforementioned descriptors. Whisker Plots with statistical value were built by DataWarrior.21 Additionally, a between-group comparison was carried out using the Mann− Whitney test. PCA has the disadvantage of inefficiently using available space, where descriptor dimensions can be missed.21 Finally, the relationship between computed molecular properties and nature types was mined by QuantMiner, which is based on a genetic algorithm to reveal association rules by optimizing both the support and the confidence.36 Murcko Scaffolds Analysis. From all screened compounds, scaffolds representing the core structures of compounds were computed according to Murcko37 in DataWarrior and analyzed using two independent methods: PCA, and 2D-Rubber Band Scaling (2D-RBS).21 The SkelSpheres descriptor was selected to build 2D-RBS plots of all scaffolds, while FragFP was selected as the descriptor to build PCA plots. Marker colors and shapes represented different nature types, respectively. Target Prediction and Enrichment Calculation. The compounds were subjected to a target prediction algorithm ̈ Bayes profiling.23,38 This model was comprising Bernoulli Naive constructed through the assimilation of over 195 million bioactivity data points deposited in the ChEMBL39 and PubChem40 repositories. Targets were predicted on a per compound bases and annotated with a full set of pathways from NCBI BioSystems.41 Enrichment calculations were performed by normalizing frequencies of the target prediction/pathway annotation of each medicinal subclass of compounds to a background of 10 000
sets of compounds randomly sampled from PubChem40 and ZINC.42 The calculation of the enrichment scores was performed as follows:43,44 1. Estimation Score. The Estimation Score is based on the number of times the frequency of the predicted targets/pathway in the random samples of the background data set is larger or equal to the frequency of predicted target/pathways in the test data set. The absolute frequency (C in eq 1) is divided by the total number of random sample sets (in this case 10 000), producing a value between 0 (enriched) and 1 (random). estimation score =
C 10 000
(1)
2. Prediction Ratio. The Prediction ratio for a given target/ pathway is calculated by comparing the frequency of prediction/ annotation in the test set (Ft) and the frequency of prediction/ annotation in the background distribution (Fb), taking into consideration the total number of predictions/annotations in each set (N). prediction ratio =
Ft /Nt Fb/Nb
(2)
3. Average Ratio. The average ratio is calculated as the average for a series of Prediction ratios for each random data set (Ri), for each predicted target in the test data set. average ratio =
1 n
n
∑ i=1
Ri Ft
(3)
In this study, enriched targets/pathways were considered if they show an estimation score equal to 0.00, and filtering via 471
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling
Figure 4. Analysis of molecular descriptors. (a) PCA of molecular descriptors shows compounds with different nature tend to distribute separately, reveals that the first and second principal components explained 35.9% and 15.2% of the total variance, respectively. (b) 2D-Plot of clogP and polar surface area shows the compounds associated with neutral nature are distributed in the upper right field separated from the other compounds. (c, d, e) Whisker plots of absolute weight, polar surface area, and clogP. (c) shows that compounds associated with hot nature are significantly lighter (Mann− Whitney test, P < 0.001) than the compounds associated with cold natures, which are in turn also significantly lighter than the neutral ones (Mann− Whitney test, P < 0.001). Panel d shows that compounds associated with neutral nature have a higher value than compounds associated with cold nature (Mann−Whitney test, P < 0.001) and compounds associated with cold nature have a higher polar surface area than compounds associated with hot nature (Mann−Whitney test, P < 0.001). Panel e presents lower value of clogP in compounds associated with cold nature as compared to compounds associated with hot nature and compounds associated with neutral nature (Mann−Whitney test, P < 0.001), there was no obvious differences between the hot group and neutral group, at the confidence level considered in the analysis (Mann−Whitney test, P = 0.055). (f, g, h) Distribution plots of absolute weight, polar surface area, and clogP. Plots c−h revealed the values of some molecular descriptors, like absolute weight, polar surface area, and clogP are obvious deferent from nature type.
together as clusters, dramatically improving the possibility to navigate and interpret enrichment results. Top three diseases associated with each targets were obtained from DisGeNET50 and classified based on “International Statistical Classification of Diseases and Related Health Problems 10th Revision: Version 2016”.51 Top ten enriched GO pathways across the three nature groups were inspected with regard to their ability to explain the MoA of the compounds classified in the nature group. QuickGo performs annotation of enriched GO pathways.52 QuickGO not only filters annotations to a specific set of GO terms, but also allows the reduction to GO slims (a lighter version of GO terms concerned with only the broad functional classification), view them as a graph and use them to construct a GO slim, in which nucleus terms (the terms of pathways input and colored by different colors) compared with other GO terms.52
descending average ratio was used to further discriminate for important targets. Target and Pathway Analysis and Annotation. Targets were classified by gene ontology (GO), molecular function, and PanthER Protein Class with the PanthER classification system.45 To improve the mechanistic understanding of the MoA, the function annotation information were extracted from UniProt46 and analysis based on the Anatomical Therapeutic Chemical Classification System (ATC) and literature search.47 The Enrichment Map was used to visualize the results of gene-sets enrichment as a network by the Enrichment Map Cytoscape Plugin.48 It operated on Gene-Sets Enrichment function Analysis (GSEA) results by DAVID.49 Only gene-sets passing conservative significance thresholds (p-value < 0.005, false discovery rate (FDR) < 5%)48 were selected for display in the Enrichment Map. Nodes represent gene-sets and edges represent mutual overlap; in this way, highly redundant gene-sets were grouped 472
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling
Figure 5. Chemical space visualizations of scaffolds in the TCM data set analyzed here. (a) Represents the structure view depicting the chemical space of all scaffolds. Similar neighbors are connected with a connecting line and the markers that represent the scaffolds are colored by different nature, imply that exclusives scaffolds can represent a nature group. (b) PCA visualizations of chemical space for scaffolds, numbers in the plots represent the number of compounds with this scaffold, shows scaffolds in the circles with different colors may be special for different nature based on their separation characteristics from other scaffolds. It shows most of the scaffolds of hot group have aromatic ring systems. The cold and neutral groups instead often contain aliphatic rings. Interestingly, the neutral group, eight out of the ten most frequent scaffolds contains cyclohexene moiety. This implied that the scaffolds for each nature group show both internal similarity and outer specificity difference when compared with other group.
■
RESULTS Analysis of Chemical Space of Different Nature Groups. The plot produced by 2D-Rubber Band Scaling (2DRBS)21 (Figure 3a) resulted in small clusters of compounds from the different nature groups, as opposed to single large clusters. This implies that neighboring compounds from TCMs often possess the same nature types, however that this is not true when looking at the data on the global scale. Tree view plots (Figure 3b−d) illustrate direct neighbors of three molecules in distinct clusters, as well as nature groups. This local, but not global, cluster structure shows that CM compounds with the same nature annotation indeed do possess similarity on the chemical level but also the underlying pluralistic structural character of TCM. We next analyzed CMs on the physicochemical level. A PCA plot of molecular descriptors (Figure 4a) reveals that the first and second principal components explained 35.9% and 15.2% of the total variance, respectively, and shape index, molecular weight, and polar surface area along the first principle component (X axis) and clogP, the number of aromatic nitrogens, as well as the number of basic nitrogens along the second principle component (Y axis), are observed to be the most important components to explain the variance in the data. More specifically, the first axis is largely size-related, while the second axis is mostly related to polarity and hydrophobicity of the compounds. A scatterplot (Figure 4b) of clogP and polar surface area shows that the compounds associated with neutral nature possess a higher value of clogP and polar surface area (upper right-hand quadrant) than the compounds with other nature annotations. Whisker plots of molecular weight (Figure 4c) illustrate that compounds associated with hot nature (mean value is 354.66) are significantly lighter than the compounds associated with cold nature (mean value 371.57, Mann−Whitney test, P < 0.001), which are in turn also significantly lighter than the neutral ones (mean value is 436.81, Mann−Whitney test, P < 0.001). Similarly, whisker plots of polar surface area (Figure 4d) shows that compounds associated with neutral nature have a higher value (mean value is 101.47) than compounds associated with cold nature (mean value is 74.045, Mann−Whitney test, P < 0.001), while in turn compounds associated with cold nature
have a higher polar surface area than compounds associated with hot nature (mean value is 67.615, Mann−Whitney test, P < 0.001). clogP (Figure 4e) is found to be lower in compounds associated with cold nature as compared to both compounds associated with hot nature and compounds associated with neutral nature (mean clogP of 2.67, 2.94, and 3.17, respectively; Mann−Whitney test, P < 0.001), while there was no significant difference between the hot group and neutral group at the confidence level considered in the analysis (Mann−Whitney test, P = 0.055). The distribution curves (Figure 4f, g, h) of molecular weight, polar surface area, and clogP clearly illustrate those internature differences encountered in TCM. We next derived rule-based classifications of nature groups using QuantMiner, yielding 24 rules (Table S1; support >1.0%, confidence >70%).36 More specifically, this comprises ten rules related to compounds associated with cold nature, 13 for compounds associated with hot nature, and just one for compounds associated with neutral nature. The number of aromatic nitrogens from two to four, the number of basic nitrogens and alkyl-amines equals to two related to compounds associated with cold nature with a confidence of 100%. Similarly, it is found that from the associatiation rules (Table S1) clogP in the interval [1.8186; 2.0187] is related to cold nature with confidence of 76.62%, whereas clogP in [3.5878; 3.6618] is related to hot nature with confidence of 76.62%. However, while this may be technically correct, such a narrow range of logP values that is specific to one nature class would be rather surprising in practice. Hence, this may more likely be attributed to insufficient sampling, and given larger data sets in the future it remains to be seen whether this turns out to be an artifact or not. In the next step, we analyzed the scaffold composition of different nature groups in TCM. Scaffold abstraction based on the Bemis and Murcko approach37 generated 685 scaffolds from the 2012 compounds in the set. Similarity clustering of 199 scaffolds, which occurred more than three times in the set (Figure 5a), showed rather clean (but local) clusters of different nature classes, in agreement with the findings presented above for full compounds. The first two dimensions taken from PCA (Figure 5b) was used to position the scaffolds in space such that clusters of similar scaffolds can easily be detected and colored 473
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling Table 1. Top Ten Scaffolds in Each Group with Different Naturea
a
It can be seen that most scaffolds of the hot group (nine out of the top ten scaffolds) have aromatic ring systems. On the contrary, the cold groups instead often (eight out of the top ten scaffolds) contains aliphatic rings, while the neutral group mostly (eight out of the top ten scaffolds) contains cyclohexene moiety.
according to different natures. The plot shows that the first two principal components explained 37.78% of the variance in the data set. It can be seen that some scaffolds may be specific to certain nature. The top ten scaffolds in each nature groups (Table 1) possessed a relatively high frequency (>5) and (with the exception of benzene) no overlap exists. Moreover, the plot and the table show that most scaffolds of the hot group (nine out of the top ten scaffolds) have aromatic ring systems. On the contrary, the cold groups instead often (eight out of the top ten scaffolds) contained aliphatic rings while the neutral group mostly (eight out of the top ten scaffolds) contains cyclohexene moiety. The results of the scaffold analysis indicate that there are significant differences between hot, cold and neutral CMs, as in overall compound space, while the intragroup similarities are markedly higher. As we have found above, the nature of CMs is also visible at the chemical structural level when considering scaffolds. This implies that scaffolds characteristic for one nature type are also linked to the traditional nature of the herbals, though this link is still incompletely understood. Overall, we can conclude from the above analysis that hot, cold, and neutral nature of CMs are reflected also on the chemical structural level, when considering both full compounds, as well as scaffolds. Previous published reports concluded that natural compounds in the Traditional Chinese Medicine Compound Database have the highest complexity and even more diverse distributions than either drug-like compounds in the MDL Drug Data Report, as well as non-drug-like compounds from the Available Chemical Directory.53−55 Furthermore, it has been hypothesized that this complexity is responsible, and necessary, for the multitarget activity mechanism of TCMs.56 However, in all of those cases no global similarities could be observed, and CMs rather cluster on the local level (into ’chemical islands’), rather than a single, or few, large chemically coherent clusters.
Since chemical structure is the molecular basis for the activity of compounds,18 this pluralistic character of molecular structure associated with each nature type may also suggest that the MoAs of TCM nature groups would show diversity overall, but also similarity within the groups. This is what we examined next in this study. Target Prediction and Annotation. Protein targets were predicted for all compounds using in silico target prediction, a process that had been used successfully for the analysis of TCMs before (on a data set without nature annotations which we have employed in the current work).2 A major consideration when deploying these models is an assessment for the reliability of generated predictions. The target prediction protocol deployed in this study utilizes Platt scaling in Scikit-learn as an approach to address the domain of applicability for models.57 In this procedure, the predictions generated by a model have been calibrated to reflect the true probability of the respective activity label and hence a kind of confidence in the prediction. A desired threshold for confidence in the models is specified at runtime (in this study it was 60% confidence) and targets are only assigned an active label upon surpassing this threshold. Overall, target prediction yielded 147 enriched targets, and the Venn diagram in Figure S2a shows the overlap of enriched targets across the three nature groups, which comprise 46 targets, while in addition each group has specific unique targets, namely, 16 for both the cold group and the hot group, and 37 for the neutral group. When compared to a background distribution, the enriched predicted calculation resulted in 57 enriched targets in three different nature groups (Table S2). Only one target (3-beta-hydroxysteroid-delta (8), delta (7)-isomerase) is shared between the hot and neutral group, suggesting there are obvious differences between the three groups. To improve the mechanistic understanding of the MoA, the enrichment map was used to 474
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling
Figure 6. Gene sets map and disease classification of enriched targets in the three classes of compounds associated with hot, cold, and neutral nature vs background. Node size represents the gene set size and edge thickness represents the degree of overlap between two gene sets. The Gene sets map resulted in 161 total gene sets significantly enriched, out of which 8 were in the neutral group, 71 in the hot group, and 82 in the cold group. In the hot group, there were 13 gene sets related to protein, 11 associated with phosphate metabolic processes, 10 associated with binding, and 8 associated with positive regulation. In the neutral group, there were 3 gene sets for both melanocortin and peptide binding. These results revealed that different targets of compounds from TCM related to different nature types show overall rather different effects and regulations in biological processes. However, because of the significant diversity in all classes an overall analysis remains difficult, and in practice often considering individual herbs and treatments appears to be required for detailed understanding.
associated with different biological functions and require further annotation. Functional annotations (Table S3) from UniProt,46 classified by Anatomical Therapeutic Chemical (ATC) Classification,47 show that each CM nature group has very different biological roles in the following classes, based on their in silico predicted target annotations: (1) In the ATC class “alimentary tract and metabolism”, pathway annotations based on the predicted targets of the cold group frequently highlight “reversible hydration of carbon dioxide” (carbonic anhydrase 1, 2) and “detoxification” (liver carboxylesterase 1) effects, which is believed to be connected with the heat-clearing and detoxification function of some cold
visualize the results of gene-sets enrichment as a network (Figure 6) based on the 57 enriched targets. The enrichment map (Figure 6) displays 161 gene-sets, 8 of which are significantly enriched in the neutral group, 71 in the hot group, and 82 in the cold group.48 For the cold group, there were 16 gene-sets associated with negative regulation, nine for histone deacetylase, seven for chromatin, seven for muscle, and five for transcription. In the hot group, there were 13 gene-sets associated with protein, 11 for phosphate metabolic process, 10 for binding, and 8 for positive regulation. In the neutral group, there were three for both melanocortin and peptide. These findings indicate that the different nature groups are fequently 475
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling
Figure 7. PanthER function groups of enriched targets vs background data sets. Targets are categorized into multiple different functional groups; including molecular function, biological process, celleular component, and protein class. This shows that many differences between three groups. This figure shows that targets were categorized into multiple different groups, including molecular function, biological process, cellular component, and protein class, and implied many differences between the three groups. For molecular function, the neutral group showed high frequency in receptor activity and catalytic activity, while hot and cold group also showed high frequency in binding except catalytic activity. Enzyme regulator activity and translation regulator activity were unique for hot nature group. For biological process, neutral nature group showed high percent in cellular process and metabolic process; cold nature group showed higher frequency in cellular component organization or biogenesis, apoptotic process than other two groups, and hot nature group showed unique frequency in reproduction, locomotion, and biological adhesion. For cellular component, neutral nature group only belongs to cell part, cold nature group showed high frequency in extracellular regions, and hot nature group scattered in several locations with moderate frequency.
type”60 and “tyrosine-protein phosphatase non-receptor type 2”61 are able to affect the intestinal sugar absorption and glucose homeostasis, and “3-beta-hydroxysteroid-delta(8), delta(7)isomerase”62 and “tumor necrosis factor”63 can regulate
medicines, for example Hedyotic dif f usa (baihua sheshecao), which was reported as body toxin-removing.58 On the other hand in the hot nature group, “casein kinase II subunit beta” is able to “regulate metabolic pathways”,59 while “protein kinase C beta 476
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling
Figure 8. Disease classification of enriched targets vs background data sets. Color intensity represents predicted percent (the numbers showed in the heatmap) of disease classification for each nature group, and shows the highest frequency in all three groups. In addition, the three groups show moderate differences in other disease classification. The neoplasms type showed the highest frequency in all three groups without very obvious difference between groups. In addition, the three groups showed moderate differences in seconder position of disease classification. Mental and behavioural disorders showed 14.29% in cold group. Both endocrine, nutritional, and metabolic diseases and diseases of the circulatory system showed 9.64% in hot group. While congenital malformations, deformations, and chromosomal abnormalities showed 16.67% in neutral group. The figure shows that there is no obvious difference in the comparisons of the scale of the top three diseases in each nature group.
hot natured Chinese herbs (including Radix aconiti, Lateralis preparata, Cortex cinnamomi, Herba epimedii, Semen cuscutae, Rhizoma polygonati, Fructus psoraleae, and Radix rehmanniae).74 (4) As for the “nervous system” ATC class, there were obvious differences between different nature groups, since most targets of the cold group tend to terminate signal transduction at the neuromuscular junction (via acetylcholinesterase activity)75 and the action of dopamine (via activity on the sodium-dependent dopamine transporter).76 Conversely, the hot nature group represents excitatory or activating effects, like excitatory synaptic transmission (via glutamate receptor 4),77 neuron channel activation (via protein kinase C epsilon type).78 The neutral nature group of targets generally possess a function in long-term potentiation and neurotransmitter release, which leads to cognition or memory modulation (such as via calcium/ calmodulin-dependent protein kinase type II subunit alpha and Galanin receptor type 3).79,80 One study suggested the distinct abilities of Chinese herbs to regulate neural cell functions appears to be corassociated with their nature identified in TCM theory, and hence nature may be a useful guide for their utility in neural degenerative diseases, which is in agreement with our current work.81 On the CNS level, many CMs with cold nature possess sedative function, like Rhizoma anemarrhenae, Rhizoma coptidis, and Cortex moutan Radix.82 On the contrary, hot CMs are believed to have excitatory functions. As an example, securinine, an alkaloid isolated from the leaves of Securinega suf f ruticosa (with hot nature), was demonstrated to be capable of exciting the central nervous system and antagonizing the γ-aminobutyric acid (GABA) receptor.83 Overall, other classes showed less pronounced differences between nature groups, which are hence only summarized here. In “musculoskeletal system”, the cold nature group mostly effect “bone resorption and osteoclast differentiation”, “muscle maturation”, and “smooth muscle cell contractility”, while the hot nature group “functions as aggrecanase to cleave aggrecan, a major proteoglycan of cartilage” and “induces the expression of PERM1 in the skeletal muscle” as enriched processes. This implies that both cold and hot groups can affect the formation and restoration of the Musculoskeletal system but in different ways. The hot nature group showed function of “normal hematopoiesis” in “blood and blood forming organs”. In
cholesterol biosynthesis and lipid metabolism. In this relationship, it was shown before that herbs with hot property, such as Radix aconiti lateralis preparata and Rhizoma zingiberis, are able to improve the energy metabolism in rats by influencing the metabolic processes involving sugars, lipids, and amino acids by regulating metabolism-related gene expression.64 The neutral nature group “controls melanogenesis” and regulates body weight. A natural product bis(4-hydroxybenzyl)sulfide, isolated from the Chinese herb with neutral nature, Gastrodia elata, indeed proved to have the ability of diminishing human melanin synthesis.65 (2) In the ATC class “cardiovascular system”, the cold nature group can regulate angiogenesis and affect both vascular endothelial cell function and heart development. Indeed, it has been reported that the formulation of a pair of herbals of cold nature, namely, Danshen−Gegen decoction, has proliferative effect on myocardium cells via the MAPK and insulin signaling pathways.66 The hot nature group possesses the function of “cardioprotection from ischemia”, which has previously been established by the hot nature medicines, namely, Herba Epimedii and Fructus Psoraleae, which is functionally attributed at least partly to an increase in coronary blood flow in the heart.67 (3) For the top-level ATC class of “genito-urinary system and sex hormones”, the cold nature group, “testosterone 17-betadehydrogenase 3”68 and “retinoic acid receptor alpha”,69 are enriched and involved in the reduction of androstenedione to testosterone and regulating retinoic acid-induced germ cell development. In the hot nature group, “lymphokine-activated killer T-cell-originated protein kinase”70,71 support testicular functions, while the “prostaglandin F2-alpha receptor”72 initiates parturition in ovarian luteal cells and thus induces luteolysis. According to TCM, many CMs with hot nature are believed to have the function of warming the kidney to invigorate yang, that is, to improve fertility and sexual function, of which Epimedium herbs (icariin) and Pilose antler (of a young stag) are examples. The hot herb Astragalus membranaceus was in one study reported to stimulate a 1.4-fold increase in washed and unwashed sperm motility.73 A related study in 9 day old female SD rats disturbed with exogenous androgen reported a reduction of androgen levels and induction of ovulation through regulation the sex gland axis and adrenal gland activity upon oral administration of 477
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling “antineoplastic and immunomodulating agents”, most predicted targets of the three groups show tumor suppression and immunoregulation effects, and not very pronounced differences between nature groups. We next compared PanthER45 groups of enriched targets, compared to background data sets. Results are displayed in Figure 7 and show that targets were categorized into multiple different groups, which include “molecular function”, “biological process”, “cellular component”, and “protein class”, with many apparent differences between the three nature classes. For molecular function, the neutral group showed high frequency in receptor activity and catalytic activity, while the hot and cold nature groups, as opposed to the neural nature class, also showed high frequency in binding (except catalytic activity). Enzyme regulator activity and translation regulator activity were uniquely associated with the hot nature group. For biological processes, the neutral nature group showed a high percentage modulating cellular processes and metabolic processes, while the cold nature group showed higher association with cellular component organization or biogenesis and apoptotic processes than the other two groups. On the other hand, the hot nature group showed unique frequency in reproduction, locomotion, and biological adhesion. For the cellular component, the neutral nature group only seems to target cellular components, while the cold nature group was more associated with activities in the extracellular regions, and the hot nature group being associated with activities across several locations. We next analyzed the diseases associated with those targets in each group. The resulting heatmap, shown in Figure 8, compares the involvement of classifications of three top scored diseases in each nature group based on DisGeNET.50 “Neoplasms” showed the highest frequency in all three groups, with this result being consistent with the result of the functional annotation (Table S4), where most of compounds show tumor suppression and immunoregulation effects without very obvious differences between groups. However, the three nature groups showed differences in the second position of annotated disease classifications; “mental and behavioural disorders” showed a frequency of 14.29% in the cold group, while both “endocrine, nutritional, and metabolic diseases” and “diseases of the circulatory system” showed a prevalence of 9.64% in the hot nature group, and “congenital malformations, deformations, and chromosomal abnormalities” occurred at a frequency of 16.67% in the neutral group. According to the principles of TCM, “treating different diseases with the same therapeutic principle” (i.e., “same treatment for different disease”84), multiple diseases might share the same cold or heat “syndrome” (also called “pattern”) and be treated by the medicines of the same nature, whereas one disease might display several different syndromes and be treated by multiple medicines with different nature. This means that same diseases with different syndromes may have different gene expression and may be the reason that a specific biological interpretation cannot be derived from the disease enrichment analysis. These results revealed that different targets associated with different nature types show different effects on biological processes. In relative terms, however, those differences in the top three diseases in each nature group are relatively small, which is consistent with the pluralistic character of TCMs and “treating different diseases with the same therapeutic principle and using the same treatment for different diseases”.84 Enriched Pathways and Annotation. The total number of enriched pathways is 4452, of which 1947 are unique pathways.
Among the unique pathways, 430 were enriched among the cold nature group, 226 in the hot nature group and 991 in the neutral nature group. The top ten enriched GO pathways across the three nature groups are shown in Table 2, with the annotation of the top ten enriched GO pathways in each nature group conducted using QuickGo.85 The top enriched pathway annotations for the cold nature type (Table 2, Figure S3, the nucleus term colored by different colors) revealed the following enrichments: (1) Calcitriol receptor activity (including: bile acid receptor activity, bile acid signaling pathway). (2) Positive regulation of vitamin D 24-hydroxylase activity (including: regulation of vitamin D 24-hydroxylase activity). This is a process that increases the rate, frequency,or extent of vitamin D 24-hydroxylase activities. (3) Positive regulation of apoptotic processes involved in mammary gland involution. The calcitriol (1,25-dihydroxy-vitamin D3) receptor (VDR) is a nuclear protein of the erbA superfamily that regulates gene expression in a ligand-dependent manner, interaction of the VDR with its response elements produces bioactive proteins and promotes the physiological actions of calcitriol.86 Optimal vitamin D intake and its status are not only important for bone and calcium-phosphate metabolism, but also for overall health and well-being.87 This indicates that cold CMs might affect bone resorption and bone formation, for which indeed some evidence exists: A review highlights the antiosteoporotic potential of the cold nature herb, Salvia miltiorrhiza in clinical applications.88 Moreover, the cold CMs might regulate immunity, autoimmunity, cardiovascular disease, cancer, fertility, pregnancy, dementia, and mortality by affecting vitamin D metabolism either directly or indirectly.87 The positive regulation of apoptotic processes involved in mammary gland involution may lead to cytotoxicity in breast cancer cells89 after application of CMs from this nature class. Evidence can be found here, since previous research revealed that berberine, a compound from the cold herb Rhizoma Coptidis, induced cytotoxicity in breast cancer cells involving reactive oxygen species (ROS) generation and apoptotic processes.90 The top enriched GO pathways annotations for the hot nature type (Table 2, Figure S4) showed the following most significant signals: (1) Primary spermatocyte growth. The phase of growth and gene expression that male germ cells undergo as they enter the spermatocyte stage. The cells grow in volume and transcribe most of the gene products needed for the morphological events that follow meiosis. This process is also part of sexual reproduction. (2) Androgen binding. This process relates interacting selectively and noncovalently with any androgen, male sex hormones. (3) Starch metabolic process. Starch is synthesized as a temporary storage form of carbon and can be catabolized to produce sucrose. This process is a cellular glucan metabolic process. (4) Maltose alpha-glucosidase activity and glucan 1,4-alpha-glucosidase activity (including: maltose metabolic process, oligosaccharide metabolic process, and amylase activity). These regulations are highly consistent with the annotation of predicted targets above, and can be grouped into two parts: the first is related to the function of hot nature CMs in improving fertility and sexual function; the second is involving processes related to energy metabolism and processes affecting body temperature. Previous research reported that the cold and hot natures of CMs may be reflected in an ethological way for the changes of animal temperature tropism, which might be internally regulated by the body’s energy metabolism.8 Other studies also proved that herbs with hot properties, such as Rhizoma zingiberis, could improve the energy metabolism in rats, 478
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling
The top enriched pathway annotations for the neutral nature type (Table 2, Figure S5) include the following: (1) Retinoic acid binding and mRNA 5″-UTR binding. These processes may relate to binding such as ion binding, small molecule binding and compound binding. (2) Translation regulator activity, nucleic acid binding. (3) Negative regulation of granulocyte differentiation. This stops, prevents, or reduces the frequency, rate or extent of granulocyte differentiation. (4) Chondroblast differentiation. The process in which a mesenchymal cell acquires specialized structural and functional features of a chondroblast. Differentiation includes the processes involved in commitment of a cell to a chondroblast fate. A chondroblast is a precursor cell to chondrocytes. This is associated with connective tissue development. (5) Sertoli cell fate commitment. The process in which the cellular identity of sertoli cells is acquired and determined. This is part of sex differentiation and related the establishment of the sex of an organism by physical differentiation. The enrichments observed for this nature type imply that neutral nature primarily shows an effect on cellular function and cellular processes as well as sex differentiation. Ginkgo biloba, a herb with neutral nature, has hypoxia-protective effects which are believed to be associated with its antioxidant activity, a property that has been shown to protect key cellular processes and structures, such as Na/K-ATPase function, mitochondrial ATP synthesis.91 Research on CMs related to sexual differentiation remains scarce, however, except for one report that the mother of a 20-year old man diagnosed with 46 XX male sex reversal syndrome had taken the traditional tocolytic Chinese medicine (the detail of the CMs were not mentioned in the paper) to prevent spontaneous abortion during pregnancy.92 These results shows that the different nature types indeed seem to affect different biological processes based on the pluralistic character of molecular structure. compounds associated with cold nature had a lower clogP and contain more aliphatic rings than the other groups and were found to control detoxification, heat-clearing, heart development processes and have sedative function, associated with mental and behavioural disorders. Compounds associated with hot nature were on average of lower molecular weight, these had more aromatic ring systems than compounds in the other groups and seemed to control body temperature, have cardioprotective functions, improve fertility and sexual function, had excitatory or activating effects, and were associated with endocrine, nutritional, and metabolic diseases and diseases of the circulatory system. Compounds associated with neutral nature had a higher polar surface area and contained more cyclohexene moietys than compounds in other groups and seemed to be related to memory function, which implies that compounds with this nature may be useful in treating neural degenerative disease. All these represents a common mechanism underlying the traditional nature theory of TCM.
Table 2. Top Ten Enriched GO Pathways Across 3 Nature Groupsa nature cold
hot
neutral
name
external id
average ratio
1. bile acid receptor activity 2. bile acid signaling pathway 3. regulation of apoptotic process involved in morphogenesis 4. positive regulation of apoptotic process involved in morphogenesis 5. calcitriol receptor activity 6. regulation of vitamin D 24hydroxylase activity 7. positive regulation of vitamin D 24-hydroxylase activity 8. regulation of mammary gland involution 9. positive regulation of mammary gland involution 10. positive regulation of apoptotic process involved in mammary gland involution 1. primary spermatocyte growth 2. retinoic acid receptor activity 3. androgen binding 4. negative regulation of cartilage development 5. oligosaccharide metabolic process 6. amylase activity 7. maltose alpha-glucosidase activity 8. glucan 1,4-alpha-glucosidase activity 9. maltose metabolic process 10. starch metabolic process 1. translation repressor activity 2. regulation of granulocyte differentiation 3. negative regulation of granulocyte differentiation 4. translation regulator activity 5. Mrna 5−UTR binding 6. sertoli cell fate commitment 7. chondroblast differentiation 8. translation regulator activity, nucleic acid binding 9. translation repressor activity, nucleic acid binding 10. retinoic acid binding
GO:0038181 GO:0038183 GO:1902337
0.025538012 0.025538012 0.025647059
GO:1902339
0.025647059
GO:0008434 GO:0010979
0.025647059 0.025647059
GO:0010980
0.025647059
GO:1903519
0.025647059
GO:1903521
0.025647059
GO:0060058
0.025647059
GO:0007285 GO:0003708 GO:0005497 GO:0061037
0.038139623 0.051210526 0.054739645 0.056168831
GO:0009311
0.061581395
GO:0016160 GO:0032450
0.061581395 0.061581395
GO:0004339
0.061581395
GO:0000023 GO:0005982 GO:0030371 GO:0030852
0.061581395 0.061581395 0.005290323 0.005290323
GO:0030853
0.005290323
GO:0045182 GO:0048027 GO:0060010 GO:0060591 GO:0090079
0.005290323 0.005290323 0.005290323 0.005290323 0.005290323
GO:0000900
0.005290323
GO:0001972
0.005290323
a
e_score = 0, query_hits > 10% of the number of each group. It can be seen that the different nature types indeed seem to affect different biological processes. Compounds associated with cold nature were found to have calcitriol receptor activity (including bile acid receptor activity, bile acid signaling pathway), positive regulation of vitamin D 24-hydroxylase activity (including regulation of vitamin D 24hydroxylase activity), and positive regulation of apoptotic processes involved in mammary gland involution. Compounds associated with hot nature seemed to improve fertility and sexual function and be involved in processes related to energy metabolism and processes affecting body temperature. Compounds associated with neutral nature shows an effect on cellular function and cellular processes, as well as sex differentiation.
■
DISCUSSION TCM understands the human body using system discrimination, with emphasis on the multiple mechanisms and multiple treatments.93 In system discrimination, the human body is identified as closely related systems that form an integrated network. The different parts of the system are connected by the meridian, blood and vital energy flow. For example, the heart system consists of the heart as a center, together with blood, blood vessel, mind, tongue, and the small intestine.93,94 Pathologically, TCM does not focus on a specific pathogen nor on pathological changes in a specific organ but aims to identify
through influencing the metabolic processes of sugars, lipids, and amino acids.64 This body temperature regulation is one of the most important function of medicines with different natures. 479
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling
there are still limitations. First, only a fraction of the entire human proteome is available in the in silico target prediction tool used for effect annotation (currently around 1080 protein targets). Thus, extending the biological space of the in silico target prediction algorithm can likely provide a more comprehensive overview of targets that are involved in the nature effects. Second, natural compounds only represent approximately 3.85% of the total compounds available.98 Thus, the limited coverage of chemical space may affect and limit the scope of this study. Despite these limitations, our study has overall been able to describe the putative MoA of compounds from the hot, cold and neutral nature classes in TCM. With the global and parallel analysis of the chemical as well as bioactivity space, this study revealed the similarity and the differences between them in an integrated fashion. Hence, this analysis not only provides a better insight of the MoAs of traditional cold, hot, and neutral nature but also holds great potential for bridging the gap between complex functions of the human body and TCMs used for its treatment.
the holistic disturbances among the self-controlled systems by analyzing all symptoms and signs.93 TCM hence places an emphasis on the dynamic changes in any parts and any connections in the self-controlled system, and regards as disorders any disturbance found in any part of the self-controlled system.93 The therapeutic mechanism in TCM focuses on enhancing the human body’s resistance to diseases and improving the interconnections among self-controlled systems by using different therapeutic methods, such as mind-spiritual methods (such as Qigong, Taiji boxing), natural methods (acupuncture, moxibustion, herbal formulas consist of multiple CMs with multiple constituents).82,95,96 This pluralistic character is considered to be one of the reasons that hinder the translation of TCM by using more “Western” research methods, which are based on monism, while TCM uses a holistic approach summarized above. The effect of this can also be seen in this study of the cold, hot, and neutral nature of compounds.97 In the present study, we characterized TCMs of cold, hot, and neutral nature using different annotations in chemical, protein target, pathway, and disease space and aimed to identify patterns within, as well as among, the different nature groups. The chemical space analysis revealed that compounds with the same nature do possess structural similarity, which we have characterized in different ways. Compounds associated with hot nature are significantly lighter in molecular weight than compounds with cold and neutral nature. Similarly, compounds associated with neutral nature have a higher polar surface area than compounds associated with cold nature, while compounds associated with cold nature have a higher polar surface area than compounds associated with hot nature. clogP was found to be lower in compounds associated with cold nature as compared to both hot and compounds associated with neutral nature (between which no characteristic difference has been found). On the scaffold level, most of the scaffolds of the hot group (nine out of the top ten scaffolds) have aromatic ring systems, to the contrary the cold groups instead often (eight out of the top ten scaffolds) contain aliphatic rings, while the neutral group mostly (eight out of the top ten scaffolds) contains single double bonded rings. We were able to find patterns in the ways CMs modulate biology in characteristic ways, by projecting the nature classes into bioactivity space. Here, the cold nature type was linked to detoxification effects, angiogenesis regulation, vascular endothelial cell function and heart development, showing blocking and termination effects in the nervous system category, antiosteoporotic potential through calcium phosphate metabolism, and cytotoxicity by regulating apoptotic processes of the mammary gland. The hot nature type involved functions on the regulation of body temperature through energy metabolism of glucan metabolic processes, sexual reproduction, cardioprotection from ischemia, and excitatory, activating, and promoting effects in the nervous system category. The neutral nature type was closely linked to controlling melanogenesis, effects on cellular function, cognition and memory, as well as sex differentiation. In many cases those links could be substantiated by literature. When linking nature groups to diseases, no obvious differences could be observed, which can be linked to the TCM theory of treating different diseases with the same therapeutic principle and using the same treatment for different diseases, depending on the overall systems picture present. Although this study has successfully rationalized TCM nature in terms of molecular structures, protein targets, and pathways,
■
ASSOCIATED CONTENT
S Supporting Information *
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jcim.6b00725. Association rules of descriptors (support >1.0%, confidence >70%), enriched targets vs background data sets, part of the target (gene) function annotation, top three targets from each group and compounds from different herbs supported by published literature, data set assembly and screening, enriched target prediction and PanthER class of all enriched targets, GO Slims and GO Term comparison of top ten enriched predicted pathways associated with cold nature type, GO Slims and GO Term comparison of top ten enriched predicted pathways associated with hot nature type, and GO Slims and GO Term comparison of top ten enriched Predicted pathways associated with neutral nature type (PDF)
■
AUTHOR INFORMATION
Corresponding Authors
*E-mail:
[email protected]. *E-mail:
[email protected]. ORCID
Xianjun Fu: 0000-0002-9399-4635 Andreas Bender: 0000-0002-6683-7546 Author Contributions
X.F. dedicated a substantial contribution to the conception and design of the project, analysis and interpretation of the data, and wrote this manuscript. L.H.W. developed the target prediction protocols, predicted the targets and pathways of compounds and helped write this manuscript. X.L., J.L., Y.Z., and Y.S. contributed to the data set the data preparation. H.Y., S.Z.M.Z., and A.Z. helped analysis and interpretation of data. A.B. and Z.W. conceived the main theme on which the work is performed and ensured that the scientific aspect of the study is rationally valid. All authors have read and approved the final manuscript. Notes
The authors declare no competing financial interest. 480
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling
■
Medicines Based on a Self-Organizing Map (SOM). Mol. Inf. 2016, 35, 109. (16) Hao, E.-W.; Deng, J.-G.; Du, Z.-C.; Yan, K.; Zheng, Z.-W.; Wang, Q.; Huang, L.-Z.; Bao, C.-H.; Deng, X.-Q.; Lu, X.-Y.; Tang, Z.-L. Experimental study on two-way application of traditional Chinese medicines capable of promoting blood circulation and removing blood stasis with neutral property in cold and hot blood stasis syndrome. Zhongguo Zhongyao Zazhi 2012, 37, 3302−3306. (17) Xiao, X.; Wang, J.; Zhao, Y.; Wang, Y.; Xiao, P. Thermodynamic outlook and practice of Chinese medicinal nature. Zhongguo Zhongyao Zazhi 2010, 35, 2207−2213. (18) Levy, C. W.; Roujeinikova, a; Sedelnikova, S.; Baker, P. J.; Stuitje, a R.; Slabas, a R.; Rice, D. W.; Rafferty, J. B. Molecular Basis of Triclosan Activity. Nature 1999, 398, 383−384. (19) Xianjun, F.; Peng, W.; Zhenguo, W. Hypothesis on Building of Chemical Constituent Element System of Cold-Heat Nature Based on Study of “Nature−Structure Relationship” of Traditional Chinese Medicine [J]. World Sci. Technol. (Modernization Tradit. Chinese Med. Mater. Medica) 2011, 13 (5), 919−924. (20) Estrada, E. Characterization of 3D Molecular Structure. Chem. Phys. Lett. 2000, 319, 713−718. (21) Sander, T.; Freyss, J.; von Korff, M.; Rufener, C. DataWarrior: An Open-Source Program For Chemistry Aware Data Visualization And Analysis. J. Chem. Inf. Model. 2015, 55, 460−473. (22) Salleb-Aouissi, A.; Vrain, C.; Nortet, C. QuantMiner: A Genetic Algorithm for Mining Quantitative Association Rules. IJCAI Int. Jt. Conf. Artif. Intell. 2007, 1035−1040. (23) Mervin, L. H.; Afzal, A. M.; Drakakis, G.; Lewis, R.; Engkvist, O.; Bender, A. Target Prediction Utilising Negative Bioactivity Data Covering Large Chemical Space. J. Cheminf. 2015, 7, 51. (24) Xu, W.; Lucke, A. J.; Fairlie, D. P. Comparing Sixteen Scoring Functions for Predicting Biological Activities of Ligands for Protein Targets. J. Mol. Graphics Modell. 2015, 57, 76−88. (25) Koutsoukas, A.; Simms, B.; Kirchmair, J.; Bond, P. J.; Whitmore, A. V.; Zimmer, S.; Young, M. P.; Jenkins, J. L.; Glick, M.; Glen, R. C.; Bender, A. From in Silico Target Prediction to Multi-Target Drug Design: Current Databases, Methods and Applications. J. Proteomics 2011, 74, 2554−2574. (26) Prathipati, P.; Mizuguchi, K. Systems Biology Approaches to a Rational Drug Discovery Paradigm. Curr. Top. Med. Chem. 2015, 16, 1009. (27) Zhao, M.; Zhou, Q.; Ma, W.; Wei, D.-Q. Exploring the LigandProtein Networks in Traditional Chinese Medicine: Current Databases, Methods, and Applications. Evid. Based. Complement. Alternat. Med. 2013, 2013, 806072. (28) Yan, X.; Zhou, J.; Xu, Z. Concept Design of Computer-Aided Study on Traditional Chinese Drugs. J. Chem. Inf. Comput. Sci. 1999, 39, 86−89. (29) ChemAxon. https://www.chemaxon.com/, 2015. (30) Chinese Pharmacopoeia Commission. Chinese Pharmacopoeia 2010 (Part I), 1st ed.; China Press of Traditional Chinese medicine: Beijing, 2010. (31) Editorial board of Chinese Materia Medica. Chinese Materia Medica; Science and Technology Press: Shanghai, 1999. (32) Dassault Systemes. Biovia Discovery Studio Modeling Environment, release 4.5; Dassault Systemes: San Diego, CA, 2015. (33) Metsalu, T.; Vilo, J. ClustVis: A Web Tool for Visualizing Clustering of Multivariate Data Using Principal Component Analysis and Heatmap. Nucleic Acids Res. 2015, 43, W566−570. (34) RStudio Team. RStudio: Integrated Development for R; RStudio: Boston, MA, 2015. (35) Vu, V. Q. ggbiplot: A ggplot2 based biplot, R package version 0.55. (36) Salleb-Aouissi, A.; Vrain, C.; Nortet, C.; Kong, X. R.; Rathod, V.; Cassard, D. QuantMiner for Mining Quantitative Association Rules. J. Mach. Learn. Res. 2013, 14, 3153−3157. (37) Bemis, G. W.; Murcko, M. a. The Properties of Known Drugs. 1. Molecular Frameworks. J. Med. Chem. 1996, 39, 2887−2893. (38) Koutsoukas, A.; Lowe, R.; KalantarMotamedi, Y.; Mussa, H. Y.; Klaffke, W.; Mitchell, J. B. O.; Glen, R. C.; Bender, A. In Silico Target
ACKNOWLEDGMENTS X.F., X.L., and Z.W. received funding from the National Natural Science Foundation of China (Grant No. 81473369), X.F. received funding from scholarship of Shandong Provincial Education Association for International Exchanges. L.H.M. received funding from the BBSRC and AstraZeneca. The authors would like to thank Xu Jing, Zhu Yumei, Mevrouw Marijke Van Moerbeke, Teng Jialin, Tian Sisheng, Mi Li, Zhang Fengcong, Yu Haifang, Fan Lei, Avid Afzal, Oscar Mendez Lucio, Wong Kah Keng, and Fredrik Svensson for their help and advices.
■
REFERENCES
(1) Qiu, J. When the East Meets the West: The Future of Traditional Chinese Medicine in the 21st Century. Natl. Sci. Rev. 2015, 2, 377−380. (2) Mohd Fauzi, F.; Koutsoukas, A.; Lowe, R.; Joshi, K.; Fan, T. P.; Glen, R. C.; Bender, A. Chemogenomics Approaches to Rationalizing the Mode-of-Action of Traditional Chinese and Ayurvedic Medicines. J. Chem. Inf. Model. 2013, 53, 661−673. (3) Li, W.-F.; Jiang, J.-G.; Chen, J. Chinese Medicine and Its Modernization Demands. Arch. Med. Res. 2008, 39, 246−251. (4) Fu, X.-J.; Liu, H.-B.; Wang, P.; Guan, H.-S. A Study on the Antioxidant Activity and Tissues Selective Inhibition of Lipid Peroxidation by Saponins from the Roots of Platycodon Grandiflorum. Am. J. Chin. Med. 2009, 37, 967−975. (5) Fu, X.; Wang, Z.; Qu, Y.; Wang, P.; Zhou, Y.; Yu, H. Study on the Networks Of “nature-Family-Component” of Chinese Medicinal Herbs Based on Association Rules Mining. Chin. J. Integr. Med. 2013, 19, 663− 667. (6) Liang, F.; Li, L.; Wang, M.; Niu, X.; Zhan, J.; He, X.; Yu, C.; Jiang, M.; Lu, A. Molecular Network and Chemical Fragment-Based Characteristics of Medicinal Herbs with Cold and Hot Properties from Chinese Medicine. J. Ethnopharmacol. 2013, 148, 770−779. (7) Ma, T.; Tan, C.; Zhang, H.; Wang, M.; Ding, W.; Li, S. Bridging the Gap between Traditional Chinese Medicine and Systems Biology: The Connection of Cold Syndrome and NEI Network. Mol. BioSyst. 2010, 6, 613. (8) Zhou, C.; Wang, J.; Zhang, X.; Zhao, Y.; Xia, X.; Zhao, H.; Ren, Y.; Xiao, X. Investigation of the Differences between the “COLD” and “HOT” Nature of Coptis Chinensis Franch and Its Processed Materials Based on Animal’s Temperature Tropism. Sci. China, Ser. C: Life Sci. 2009, 52, 1073−1080. (9) Li, P. P. Toward an Integrative Framework of Indigenous Research: The Geocentric Implications of Yin-Yang Balance. Asia Pacific J. Manag. 2012, 29, 849−872. (10) Jin, R.; Zhang, B.; Liu, X.-Q.; Liu, S.-M.; Liu, X.; Li, L.-Z.; Zhang, Q.; Xue, C.-M. [Study of biological performance of Chinese materia medica with either a cold or hot property based on the three-element mathematical analysis model]. Zhongxiyi Jiehe Xuebao 2011, 9, 715− 724. (11) Huang, L.; Zhu, M.; Yu, R.; Du, J.; Liu, H. Study on discrimination mode of cold and hot properties of traditional Chinese medicines based on biological effects. Zhongguo Zhong Yao Za Zhi 2014, 39, 3353−3358. (12) Zhao, H.; Zhao, Y.; Wang, J.; Li, H.; Ren, Y.; Zhou, C.; Yan, D.; Xiao, X. Expression of the Difference between the Cold (Han) and Hot (Re) Natures of Traditional Chinese Medicines (Strobal and Rhubarb) Based on the Cold/hot Plate Differentiating Assay. Sci. China, Ser. C: Life Sci. 2009, 52, 1192−1197. (13) Zhao, Y.; Jia, L.; Wang, J.; Zou, W.; Yang, H.; Xiao, X. Cold/hot Pad Differentiating Assay of Property Differences of Mahuang and Maxingshigan Decoctions. Pharm. Biol. 2016, 54, 1298. (14) Wan, H.-Y.; Kong, X.-Y.; Li, X.-M.; Zhu, H.-W.; Su, X.-H.; Lin, N. Effect of traditional Chinese medicines with different properties on thermoregulation and temperature-sensitive transient receptor potentialion channel protein of rats with yeast-induced fever. Zhongguo Zhong Yao Za Zhi 2014, 39, 3813−3818. (15) Wang, M.; Li, L.; Yu, C.; Yan, A.; Zhao, Z.; Zhang, G.; Jiang, M.; Lu, A.; Gasteiger, J. Classification of Mixtures of Chinese Herbal 481
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling Predictions: Defining a Benchmarking Data Set and Comparison of Performance of the Multiclass Naive Bayes and Parzen-Rosenblatt Window. J. Chem. Inf. Model. 2013, 53, 1957−1966. (39) Papadatos, G.; Gaulton, A.; Hersey, A.; Overington, J. P. Activity, Assay and Target Data Curation and Quality in the ChEMBL Database. J. Comput.-Aided Mol. Des. 2015, 29, 885−896. (40) Kim, S.; Thiessen, P. A.; Bolton, E. E.; Chen, J.; Fu, G.; Gindulyte, A.; Han, L.; He, J.; He, S.; Shoemaker, B. A.; Wang, J.; Yu, B.; Zhang, J.; Bryant, S. H. PubChem Substance and Compound Databases. Nucleic Acids Res. 2016, 44, D1202−D1213. (41) Geer, L. Y.; Marchler-bauer, A.; Geer, R. C.; Han, L.; He, J.; He, S.; Liu, C.; Shi, W.; Bryant, S. H. The NCBI BioSystems Database. Nucleic Acids Res. 2010, 38, D492−D496. (42) Irwin, J. J.; Sterling, T.; Mysinger, M. M.; Bolstad, E. S.; Coleman, R. G. ZINC: A Free Tool to Discover Chemistry for Biology. J. Chem. Inf. Model. 2012, 52, 1757−1768. (43) Liggi, S.; Drakakis, G.; Hendry, A. E.; Hanson, K. M.; Brewerton, S. C.; Wheeler, G. N.; Bodkin, M. J.; Evans, D. a.; Bender, A. Extensions to in Silico Bioactivity Predictions Using Pathway Annotations and Differential Pharmacology Analysis: Application to Xenopus Laevis Phenotypic Readouts. Mol. Inf. 2013, 32, 1009−1024. (44) Liggi, S.; Drakakis, G.; Koutsoukas, A.; Cortes-Ciriano, I.; Martínez-Alonso, P.; Malliavin, T. E.; Velazquez-Campoy, A.; Brewerton, S. C.; Bodkin, M. J.; Evans, D. A.; Glen, R. C.; Carrodeguas, J. A.; Bender, A. Medicinal Chemistry Extending in Silico Mechanism-of- Action Analysis by Annotating Targets with Pathways: Application to Cellular Cytotoxicity Readouts. Future Med. Chem. 2014, 6, 2029−2056. (45) Mi, H.; Muruganujan, A.; Thomas, P. D. PANTHER in 2013: Modeling the Evolution of Gene Function, and Other Gene Attributes, in the Context of Phylogenetic Trees. Nucleic Acids Res. 2013, 41, D377−D386. (46) The UniProt Consortium. UniProt: A Hub for Protein Information. Nucleic Acids Res. 2015, 43, D204−D212. (47) Liu, Z.; Guo, F.; Gu, J.; Wang, Y.; Li, Y.; Wang, D.; Lu, L.; Li, D.; He, F. Similarity-Based Prediction for Anatomical Therapeutic Chemical Classification of Drugs by Integrating Multiple Data Sources. Bioinformatics 2015, 31, 1788−1795. (48) Merico, D.; Isserlin, R.; Stueker, O.; Emili, A.; Bader, G. D. Enrichment Map: A Network-Based Method for Gene-Set Enrichment Visualization and Interpretation. PLoS One 2010, 5, e13984. (49) Huang, D. W.; Lempicki, R. A.; Sherman, B. T. Systematic and Integrative Analysis of Large Gene Lists Using DAVID Bioinformatics Resources. Nat. Protoc. 2008, 4, 44−57. (50) Pinero, J.; Queralt-Rosinach, N.; Bravo, A.; Deu-Pons, J.; BauerMehren, A.; Baron, M.; Sanz, F.; Furlong, L. I. DisGeNET: A Discovery Platform for the Dynamical Exploration of Human Diseases and Their Genes. Database 2015, 2015, bav028−bav028. (51) WHO/SDE/OE. Series P of the HEO and EH International Statistical Classification of Diseases and Related Health Problems (ICD10) in Occupational Health. World Health Org. Sustain. Dev. Health Environ. 1999, 1−36. (52) Binns, D.; Dimmer, E.; Huntley, R.; Barrell, D.; O'Donovan, C.; Apweiler, R. QuickGO: A Web-Based Tool for Gene Ontology Searching. Bioinformatics 2009, 25, 3045−3046. (53) Shen, M.; Tian, S.; Li, Y.; Li, Q.; Xu, X.; Wang, J.; Hou, T. DrugLikeness Analysis of Traditional Chinese Medicines: 1. Property Distributions of Compounds and Natural Compounds from Traditional Chinese Medicines. J. Cheminf. 2012, 4, 31. (54) Tian, S.; Li, Y.; Wang, J.; Xu, X.; Xu, L.; Wang, X.; Chen, L.; Hou, T. Drug-Likeness Analysis of Traditional Chinese Medicines: 2. Characterization of Scaffold Architectures for Drug-like Compounds, Non-Drug-like Compounds, and Natural Compounds from Traditional Chinese Medicines. J. Cheminf. 2013, 5, 5. (55) Tian, S.; Wang, J.; Li, Y.; Xu, X.; Hou, T. Drug-Likeness Analysis of Traditional Chinese Medicines: Prediction of Drug-Likeness Using Machine Learning Approaches. Mol. Pharmaceutics 2012, 9, 2875−2886. (56) Tian, S.; Li, Y.; Li, D.; Xu, X.; Wang, J.; Zhang, Q.; Hou, T. Modeling Compound−Target Interaction Network of Traditional
Chinese Medicines for Type II Diabetes Mellitus: Insight for Polypharmacology and Drug Design. J. Chem. Inf. Model. 2013, 53, 1787−1803. (57) Pedregosa, F.; Varoquaux, G.; Gramfort, A.; Michel, V.; Thirion, B.; Grisel, O.; Blondel, M.; Prettenhofer, P.; Weiss, R.; Dubourg, V.; Vanderplas, J.; Passos, A.; Cournapeau, D.; Brucher, M.; Perrot, M.; Duchesnay, É. Scikit-Learn: Machine Learning in Python. J. Mach. Learn. Res. 2012, 12, 2825−2830. (58) Shi, J.; Liu, Y.; Fang, Y.; Xu, G.; Zhai, H.; Lu, L. Traditional Chinese Medicine in Treatment of Opiate Addiction. Acta Pharmacol. Sin. 2006, 27, 1303−1308. (59) Voss, H.; Wirkner, U.; Jakobi, R. Structure of the Gene Encoding Human Casein Kinase II Subunit Beta. J. Biol. Chem. 1991, 13706− 13711. (60) Helliwell, P. a; Rumsby, M. G.; Kellett, G. L. Intestinal Sugar Absorption Is Regulated by Phosphorylation and Turnover of Protein Kinase C betaII Mediated by Phosphatidylinositol 3-Kinase- and Mammalian Target of Rapamycin-Dependent Pathways. J. Biol. Chem. 2003, 278, 28644−28650. (61) Shi, Y.; Wang, J.; Chandarlapaty, S.; Cross, J.; Thompson, C.; Rosen, N.; Jiang, X. PTEN Is a Protein Tyrosine Phosphatase for IRS1. Nat. Struct. Mol. Biol. 2014, 21, 522−527. (62) Braverman, N.; Lin, P.; Moebius, F. F.; Obie, C.; Moser, A.; Glossmann, H.; Wilcox, W. R.; Rimoin, D. L.; Smith, M.; Kratz, L.; Kelley, R. I.; Valle, D. Mutations in the Gene Encoding 3 BetaHydroxysteroid-Delta 8, Delta 7-Isomerase Cause X-Linked Dominant Conradi-Hünermann Syndrome. Nat. Genet. 1999, 22, 291−294. (63) Hotamisligil, G. S.; Shargill, N. S.; Spiegelman, B. M. Adipose Expression of Tumor Necrosis Factor-Alpha: Direct Role in ObesityLinked Insulin Resistance. Science 1993, 259, 87−91. (64) Yu, H.-Y.; Wang, S.-J.; Teng, J.-L.; Ji, X.-M.; Wu, Z.-C.; Ma, Q.-C.; Fu, X.-J. Effects of Radix Aconiti Lateralis Preparata and Rhizoma Zingiberis on Energy Metabolism and Expression of the Genes Related to Metabolism in Rats. Chin. J. Integr. Med. 2012, 18, 23−29. (65) Chen, W.-C.; Tseng, T.-S.; Hsiao, N.-W.; Lin, Y.-L.; Wen, Z.-H.; Tsai, C.-C.; Lee, Y.-C.; Lin, H.-H.; Tsai, K.-C. Discovery of Highly Potent Tyrosinase Inhibitor, T1, with Significant Anti-Melanogenesis Ability by Zebrafish in Vivo Assay and Computational Molecular Modeling. Sci. Rep. 2015, 5, 7995. (66) Fong, C. C.; Wei, F.; Chen, Y.; Yu, W. K.; Koon, C. M.; Leung, P. C.; Fung, K. P.; Lau, C. B. S.; Yang, M. Danshen-Gegen Decoction Exerts Proliferative Effect on Rat Cardiac Myoblasts H9c2 via MAPK and Insulin Pathways. J. Ethnopharmacol. 2011, 138, 60−66. (67) Patacchini, R.; Lecci, A.; Holzer, P.; Maggi, C. A. Newly Discovered Tachykinins Raise New Questions about Their Peripheral Roles and the Tachykinin Nomenclature. Trends Pharmacol. Sci. 2004, 25, 1−3. (68) Luan, X.; Liu, D.; Cao, Z.; Luo, L.; Liu, M.; Gao, M.; Zhang, X. Transcriptome Profiling Identifies Differentially Expressed Genes in Huoyan Goose Ovaries between the Laying Period and Ceased Period. PLoS One 2014, 9, e113211. (69) Doyle, T. J.; Braun, K. W.; McLean, D. J.; Wright, R. W.; Griswold, M. D.; Kim, K. H. Potential Functions of Retinoic Acid Receptor A in Sertoli Cells and Germ Cells during Spermatogenesis. Ann. N. Y. Acad. Sci. 2007, 1120, 114−130. (70) Zhao, S.; Dai, J.; Zhao, W.; Xia, F.; Zhou, Z.; Wang, W.; Gu, S.; Ying, K.; Xie, Y.; Mao, Y. PDZ-Binding Kinase Participates in Spermatogenesis. Int. J. Biochem. Cell Biol. 2001, 33, 631−636. (71) Oh, S.-M.; Zhu, F.; Cho, Y.-Y.; Lee, K. W.; Kang, B. S.; Kim, H.-G.; Zykova, T.; Bode, A. M.; Dong, Z. T-Lymphokine-Activated Killer CellOriginated Protein Kinase Functions as a Positive Regulator of c-JunNH2-Kinase 1 Signaling and H-Ras-Induced Cell Transformation. Cancer Res. 2007, 67, 5186−5194. (72) Weems, C. W.; Vincent, D. L.; Weems, Y. S. Roles of Prostaglandins (PG) F2 Alpha, E1, E2, Adenosine, Oestradiol-17 Beta, Histone-H2A and Progesterone of Conceptus, Uterine or Ovarian Origin during Early and Mid Pregnancy in the Ewe. Reprod., Fertil. Dev. 1992, 4, 289−295. 482
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483
Article
Journal of Chemical Information and Modeling
(93) Lu, A. P.; Jia, H. W.; Xiao, C.; Lu, Q. P. Theory of Traditional Chinese Medicine and Therapeutic Method of Diseases. World J. Gastroenterol. 2004, 10, 1854−1856. (94) Lukman, S.; He, Y.; Hui, S.-C. Computational Methods for Traditional Chinese Medicine: A Survey. Comput. Methods Programs Biomed. 2007, 88, 283−294. (95) Gu, P.; Chen, H. Modern Bioinformatics Meets Traditional Chinese Medicine. Briefings Bioinf. 2014, 15, 984−1003. (96) Xue, R.; Fang, Z.; Zhang, M.; Yi, Z.; Wen, C.; Shi, T. TCMID: Traditional Chinese Medicine Integrative Database for Herb Molecular Mechanism Analysis. Nucleic Acids Res. 2013, 41, D1089−D1095. (97) Li, T. Philosophic Perspective: A Comparative Study of Traditional Chinese Medicine and Western Medicine. Asian Soc. Sci. 2011, 7, 198−201. (98) Bender, A. Databases: Compound Bioactivities Go Public. Nat. Chem. Biol. 2010, 6, 309−309.
(73) Crimmel, a S.; Conner, C. S.; Monga, M. Withered Yang: A Review of Traditional Chinese Medical Treatment of Male Infertility and Erectile Dysfunction. J. Androl. 2001, 22, 173−182. (74) Zhou, J.; Qu, F. Treating Gynaecological Disorders with Traditional Chinese Medicine: A Review. Afr. J. Tradit., Complementary Altern. Med. 2009, 6, 494−517. (75) Peng, H. B.; Xie, H.; Rossi, S. G.; Rotundo, R. L. Acetylcholinesterase Clustering at the Neuromuscular Junction Involves Perlecan and Dystroglycan. J. Cell Biol. 1999, 145, 911−921. (76) Ciliax, B. J.; Heilman, C.; Demchyshyn, L. L.; Pristupa, Z. B.; Ince, E.; Hersch, S. M.; Niznik, H. B.; Levey, A. I. The Dopamine Transporter: Immunochemical Characterization and Localization in Brain. J. Neurosci. 1995, 15, 1714−1723. (77) Trussell, L. O.; Fischbach, G. D. Glutamate Receptor Desensitization and Its Role in Synaptic Transmission. Neuron 1989, 3, 209−218. (78) Skalnikova, H.; Vodicka, P.; Pelech, S.; Motlik, J.; Gadher, S. J.; Kovarova, H. Protein Signaling Pathways in Differentiation of Neural Stem Cells. Proteomics 2008, 8, 4547−4559. (79) Ö gren, S. O.; Kuteeva, E.; Elvander-Tottie, E.; Hökfelt, T. Neuropeptides in Learning and Memory Processes with Focus on Galanin. Eur. J. Pharmacol. 2010, 626, 9−17. (80) Matthews, R. P.; Guthrie, C. R.; Wailes, L. M.; Zhao, X.; Means, A. R.; McKnight, G. S. Calcium/calmodulin-Dependent Protein Kinase Types II and IV Differentially Regulate CREB-Dependent Gene Expression. Mol. Cell. Biol. 1994, 14, 6107−6116. (81) Liu, Y.-Q.; Cheng, M.-C.; Wang, L.-X.; Zhao, N.; Xiao, H.-B.; Wang, Z.-T. Functional Analysis of Cultured Neural Cells for Evaluating Cold/cool-and Hot/warm-Natured Chinese Herbs. Am. J. Chin. Med. 2008, 36, 771−781. (82) Jiang, W.-Y. Therapeutic Wisdom in Traditional Chinese Medicine: A Perspective from Modern Science. Trends Pharmacol. Sci. 2005, 26, 558−563. (83) Beutler, J. A.; Karbon, E. W.; Brubaker, A. N.; Malik, R.; Curtis, D. R.; Enna, S. J. Securinine Alkaloids: A New Class of GABA Receptor Antagonist. Brain Res. 1985, 330, 135−140. (84) Jiang, M.; Zhang, C.; Zheng, G.; Guo, H.; Li, L.; Yang, J.; Lu, C.; Jia, W.; Lu, A. Traditional Chinese Medicine Zheng in the Era of Evidence-Based Medicine: A Literature Analysis. Evid. Based. Complement. Alternat. Med. 2012, 2012, 409568. (85) Huntley, R. P.; Sawford, T.; Mutowo-Meullenet, P.; Shypitsyna, A.; Bonilla, C.; Martin, M. J.; O’Donovan, C. The GOA Database: Gene Ontology Annotation Updates for 2015. Nucleic Acids Res. 2015, 43, D1057−D1063. (86) Patel, S. R.; Qiong Ke, H.; Vanholder, R.; Hsu, C. H. Inhibition of Nuclear Uptake of Calcitriol Receptor by Uremic Ultrafiltrate. Kidney Int. 1994, 46, 129−133. (87) Pludowski, P.; Holick, M. F.; Pilz, S.; Wagner, C. L.; Hollis, B. W.; Grant, W. B.; Shoenfeld, Y.; Lerchbaum, E.; Llewellyn, D. J.; Kienreich, K.; Soni, M. Vitamin D Effects on Musculoskeletal Health, Immunity, Autoimmunity, Cardiovascular Disease, Cancer, Fertility, Pregnancy, Dementia and mortalityA Review of Recent Evidence. Autoimmun. Rev. 2013, 12, 976−989. (88) Guo, Y.; Li, Y.; Xue, L.; Severino, R. P.; Gao, S.; Niu, J.; Qin, L. P.; Zhang, D.; Brömme, D. Salvia Miltiorrhiza: An Ancient Chinese Herbal Medicine as a Source for Anti-Osteoporotic Drugs. J. Ethnopharmacol. 2014, 155, 1401−1416. (89) Rosfjord, E. C.; Dickson, R. B. Growth Factors, Apoptosis, and Survival of Mammary Epithelial Cells. J. Mammary Gland Biol. Neoplasia 1999, 4, 229−237. (90) Patil, J. B.; Kim, J.; Jayaprakasha, G. K. Berberine Induces Apoptosis in Breast Cancer Cells (MCF-7) through MitochondrialDependent Pathway. Eur. J. Pharmacol. 2010, 645, 70−78. (91) Gertsch, J. H.; Seto, T. B.; Mor, J.; Onopa, J. Ginkgo Biloba for the Prevention of Severe Acute Mountain Sickness (AMS) Starting One Day before Rapid Ascent. High Alt. Med. Biol. 2002, 3, 29−37. (92) Wang, T.; Liu, J. H.; Yang, J.; Chen, J.; Ye, Z. Q. 46, XX Male Sex Reversal Syndrome: A Case Report and Review of the Genetic Basis. Andrologia 2009, 41, 59−62. 483
DOI: 10.1021/acs.jcim.6b00725 J. Chem. Inf. Model. 2017, 57, 468−483