Virtual screening, Docking, ADMET and System Pharmacology studies on Garcinia caged Xanthone derivatives for Anticancer activity

Caged xanthones are bioactive compounds mainly derived from the Garcinia genus. In this study, a structure-activity relationship (SAR) of caged xanthones and their derivatives for anticancer activity against different cancer cell lines such as A549, HepG2 and U251 were developed through quantitative (Q)-SAR modeling approach. The regression coefficient (r2), internal cross-validation regression coefficient (q2) and external cross-validation regression coefficient (pred_r2) of derived QSAR models were 0.87, 0.81 and 0.82, for A549, whereas, 0.87, 0.84 and 0.90, for HepG2, and 0.86, 0.83 and 0.83, for U251 respectively. These models were used to design and screened the potential caged xanthone derivatives. Further, the compounds were filtered through the rule of five, ADMET-risk and synthetic accessibility. Filtered compounds were then docked to identify the possible target binding pocket, to obtain a set of aligned ligand poses and to prioritize the predicted active compounds. The scrutinized compounds, as well as their metabolites, were evaluated for different pharmacokinetics parameters such as absorption, distribution, metabolism, excretion, and toxicity. Finally, the top hit compound 1G was analyzed by system pharmacology approaches such as gene ontology, metabolic networks, process networks, drug target network, signaling pathway maps as well as identification of off-target proteins that may cause adverse reactions.

Descriptor calculation. The descriptors were calculated by using the QSARPlus module of VLifeMDS ® .
These descriptors were grouped into three classes, namely physicochemical descriptors, alignment independent descriptors and atom-type count descriptors. Further, these major classes were categorized in several sub-classes and each individual descriptor has been calculated for each of the training set compounds. In order to select the best subset of descriptors, highly correlated descriptors were excluded. The correlation matrix for derived models was provided in Tables S7-S9.

QSAR model development.
Prior to the QSAR model development selected experimental dataset was divided into the training (Tables S1-S3) and test set (Tables S4-S6). To divide the dataset the sphere exclusion clustering method was used 25 . The validity of training and test set were compared through the UniColoumn statistics approach. Finally, the QSAR models were developed based on stepwise forward MLR method 26 .

Statistical evaluation/validation of model. Statistical validation is a very important process of robust
QSAR model development. Therefore, various types of statistical validations were procured and the best model was selected by applying these different statistical parameters. Listed below are validations performed for the tested alignment, e.g., (i) Coefficient of determination (r 2 ) should >0.7, (ii) Leave-one-out cross-validation (LOOcv) or correlation coefficient (q 2 ) should be >0.5, (iii) Correlation coefficient of external validation set (pred_r 2 ) should be >0. 5. This is required to verify the ability of derived model in order to predict the biological activity of similar molecules, (iv) Correlation metrics (rm 2 ) calculated based on the correlation between the observed and predicted response data, with or without the intercept and also by interchanging the axes. For an acceptable QSAR model, the value of ' Average rm 2 ' should be >0.5 & 'Delta rm 2 ' should be <0.2 24 , (v) Degree of freedom (Df) should be higher (higher is better), (vi) F-test for statistical significance of the model (higher is better, for the same set of descriptors and compounds), (vii) Highest q 2 value in the randomization test (best_ ran_q 2 ) should be low as compared to q 2 , (viii) Highest r 2 value in the randomization test (best_ran_r 2 ) should be low as compared to r 2 , (ix) Statistical significance parameter by randomization (alpha_test) should be <0.01, (x) Standard error of estimate (r 2 _se, q 2 _se and, pred_r 2 se) should be smaller (smaller is better), (xi) Z-score calculated by the randomization test (ZScore) should be higher (higher is better) and, (xii) Applicability Domain (AD)/Extrapolation parameter defined as '0' and '1' . The '0' indicate applicability and data are true, whereas '1' indicate the outlier 26 .
Designing of novel caged xanthone derivatives. Based on structural feature selection in terms of chemical descriptors and 3D structural components (pharmacophores) of derived QSAR models, activity controlling sites were identified for each model and accordingly novel caged xanthone derivatives were virtually designed.
Rule of five, ADMET risk screening and Synthetic accessibility assessment. Lipinski's rule of five was used for the screening of xanthone derivatives and filtered the orally bioavailable compounds 27 . Further, these filtered xanthone derivatives were evaluated for ADMET Risk parameters by using ADMET Predictor TM (Simulations Plus, USA). The overall risk was considered to be in the range of 0-24, where a lower score is preferable and show better druggability 28 . Later the synthetic accessibility was measured in terms of a score on a scale from 1 (very easy to synthesize) to 10 (complex and challenging to synthesize) by using the SYLVIA-XT 1.4 16 .
Target identification and Molecular Docking Studies. Target identification: The possible drug targets for the identified hits were explored with the help of MetaDrug TM (Thomson Reuters, USA). Through this, a list of targets is identified, such as ESR1 (Estrogen Receptor 1), GCR (glucocorticoid receptor), FGF (Fibroblast growth factors), CREBP1 (Cyclic AMP Response Element-Binding Protein) and STATs (signal transducer and activator of transcription) (Fig. S1).

Protein preparation.
To prepare the target protein, structure of identifying proteins were retrieved from the RCSB PDB database 29 . In the first step, the protein preparation protocol was used. This protocol performs tasks such as modeling missing loop regions, inserting missing atoms in incomplete residues, deleting alternate conformations and standardizing names of the atoms, protonating titratable residues, and removing water 30  Protein-ligand Docking Studies. The molecular docking and visualization studies were performed with the help of the LibDock program in Discovery Studio v3. 5 (Accelrys, USA) 31 . The LibDock is a flexible docking module. LibDock uses protein site features, referred to as hot spots, consisting of two types states (polar and apolar). The ligand poses are placed into the polar and apolar receptor interactions site. A polar hotspot is preferred by a polar ligand atom (e.g., a hydrogen bond donor or acceptor), and an apolar hotspot is preferred by an apolar atom (e.g., a carbon atom) 32 . The protocol allows the user to specify several modes for generating ligand conformations for docking. Conformer Algorithm based on Energy Screening And Recursive buildup (CAESAR) was used for generating the conformations 33 . The smart minimizer was used for in situ ligand minimization. Further, to identify definite interacting residues of the receptor with bound ligand a 2D diagram of docking was also performed. The different scoring protocol was used for the scoring functions such as Jain, Ludi, potential of mean force (-PMF) and piecewise linear potential (PLP1) to evaluate ligand binding in a receptor cavity 34 .
In silico pharmacokinetics, metabolism, and toxicity studies. Different pharmacokinetics parameters, namely, Absorption, Distribution, Metabolism, Excretion, and Toxicity were calculated. This study includes the quantitative measurement of drug-like properties, pKa, absorption, solubility, lipophilicity, bioavailability, permeability, volume of distribution, blood-brain-barrier (BBB) penetration, hepatic clearance, transporters, dermal and ocular penetration, plasma-protein binding, metabolism, drug-drug interaction, half-life, etc. A plethora of diverse enzyme families and proteins are involved in xenobiotic metabolism, including CYP450 enzymes. These enzymes were detected for predicted active xanthone derivatives. The different metabolites and the sites of metabolism for Phase I and Phase II metabolism were also calculated. The safety of the compounds is an important parameter for a successful drug. For this, the hepatotoxicity, nephrotoxicity, neurotoxicity, and chance of causing anemia etc. is calculated. We also study the effect of the compound on different liver associated enzymes such as alkaline phosphatase (ALP), gamma-glutamyl transpeptidase (GGT), serum glutamic oxaloacetic (SGOT) & serum glutamate-pyruvate transaminase (SGPT), and lactate dehydrogenase (LDH) enzymes when administered. These findings will be helpful to set dose-ranges. These studies were performed by using the ADMET Predictor TM , MetaDrug TM , MetaPrint2D, MedChem Designer TM and TOPKAT module of Discovery Studio molecular modeling software 35 .
System pharmacology study for enrichment analysis. Identification and ranking of cellular pathways and biological processes which was most influenced by candidate compound 1G were performed by MetaDrug TM tool. For this, the enrichment analysis, such as process networks, metabolic networks, and their pathway map analysis was projected. The enrichment significance was measured in terms of −log (p-value). Lower −log (p-value) means higher the relevance of an entity 26 . Signal transduction pathway exploration to identify the possible mechanism of action. In a quest to understand the possible mechanism of action of predicted top hit compound 1G, signal transduction pathway exploration study was performed and identified the pathway and the associated functional genes affected by the top hit compound through MetaDrug TM software.
Off-targets prediction for compound 1G. Numerous drugs are known for their multi-targeting activities. Thus, keeping in mind these off-target interactions, the complexity of candidate compounds for any such interference with other human protein targets was evaluated by different prediction modules of MetaDrug TM platform.

Results and Discussion
QSAR model development and its validation. In the present study, three statistical regression QSAR models were developed, so that to study the xanthones derivatives activity against three different cancer cell lines A549 (Model 1), HepG2 (Model 2), and U251 (Model 3). The study results, several models and the best model was selected based on various statistical parameters such as a square of the correlation coefficient or regression coefficient (r 2 ), and the robustness of model predictions was estimated from the cross-validated squared correlation coefficient (q 2 ). The plot of observed verses predicted activity provides an idea about how well the model was trained and how well it predicts the activity of the external test set Fig. 1A-C. These derived QSAR models are discussed below in details, and showing the relationship between in vitro experimental activity (i.e., IC 50 ) as the dependent variable and independent variables (chemical descriptors).

QSAR Model 1.
Result shows that the model 1 yielded a high activity-descriptors relationship accuracy of 87% referred by regression coefficient (r 2 = 0.87). It showed the internal (q 2 ) and external (pred_r 2 ) predictive ability of about 81% and 82% respectively. The fitness plot for the training and test set was provided in Fig. 1A. A radar plot representing the closeness between the actual and predicted activity of training and test set compounds was also summarized in Fig. S2. The equation 1 shows that the descriptors which play an important role in determining the anti-cancer activity are SdsCHE-index, MMFF_29, SssssCcount, DeltaEpsilonC; and T_2_2_1. The contributions (positive and negative) aggregate of each of the descriptors was provided in Fig. S3A. QSAR Model 2. This model conceded an activity-descriptors relationship accuracy of 87% referred by regression coefficient (r 2 = 0.87). The internal (q 2 ) and external (pred_r 2 ) predictive ability was 84% and 90% respectively. The fitness plot for the training and test set was provided in Fig. 1B. With this, a radar plot which represents the closeness between the actual and predicted activity of the compounds of training and test set were provided in Fig QSAR Model 3. The derived QSAR model generated a good activity-descriptors relationship accuracy of 86% referred by regression coefficient (r 2 = 0.86). The model exhibit internal (q 2 ) and external (pred_r 2 ) predictive ability of 83% each. The fitness plot for the training and test set was provided in (Fig. 1C). Consecutively a radar plot was provided in Fig. S5 which represents the closeness between the actual and predicted activity of training and test set compounds. The model (equation 3) showed that three descriptors were highly correlated with the biological activity, such as T_T_N_4, T_O_O_3, and SssssCE-index. All the descriptors were directly proportional to the activity and were presented through a bar graph in Fig. S3C. Validation of the developed models. Validation is a very important process for QSAR models. Therefore, various types of validations were procured and the best model was selected by applying different statistical parameters. All the models were successfully validated by using a random test set compounds (Tables S4-S6), along with other parameters such as r 2 , (LOOcv) q 2 , pred_r 2 , Df, F test, best_ran_r 2 , best_ran_q 2 , alpha_test, r 2 se, q 2 se, pred_r 2 _se, ZScore, Applicability domain etc. and the result was provided in  Table 1). The residual error plot of the respective model was shown in Fig. S6. The result shows that the developed models were robust.

Virtual designing and filtering of novel GX derivatives. Designing of compounds.
A library set of about 1000 compounds was designed by using the useful descriptors identified through the QSAR models. Out of 1000 designed xanthone derivatives, 350 compounds were designed based on structural feature selection through A549 cancer cell line specific QSAR model, 350 compounds were designed based on feature selection through HepG2 cell line specific QSAR model, and 300 derivatives were designed based on feature selection through U251 cell line specific QSAR model. Further, these compounds are predicted for its anticancer activity by using the respective models. The compound which has the IC 50 value of more than 20 µM has been excluded and the rest are carried for additional studies (Tables S10-S12).
Screening through Lipinski's Rule of Five, ADMET Risk and synthetic accessibility. The possible active compounds were further screened through Lipinski's rule of five for oral bioavailability, which screened out 300 designed compounds. Further, screened compounds were analyzed through pre-ADMET risk screening study. For this, a score based risk was calculated to identify the real ADMET problem behind the compound so that to prevent later failure of the compound. The compounds with an ADMET risk of 10 or more were excluded. The result of  candidate compounds 1G along with control compound was shown in Table 2. The result indicates that the 1G showed the risk of 7.3 in comparing to control drug topotecan, which shows a score of 2.0. The risk parameters evaluated were size, charge, water solubility, the volume of distribution, acute rat toxicity and carcinogenicity, SGOT elevation, hepatotoxicity and inhibition of 3A4 oxidation of midazolam ( Table 2). The query compounds were screened for their synthetic accessibility. To measure this, the complexity of the ring system, complexity of the molecular structure, similar to commercially available compounds, the number of stereocenters and the potential for using important synthetic reactions were independently weighted to provide a single value for the synthetic accessibility score. Those compounds which show the high score was removed. The Synthetic accessibility score of compound 1G was found to be 7.4 ( Table 2). These hydrogen bonds may be considered as a stabilizer of the complex and may be the reason for conformational stability and thus resulted in significant activity (Fig. 2B). The interactive amino acid residues were presented in Table 3. These interactions were displayed with 2D diagram and represented by different colors e.g., pink indicates electrostatic interactions; purple indicates the covalent bond, and green indicates Van der-Waals interaction. Solvent accessibility of the ligand atoms and the amino acid residues are shown in a light blue shade surrounding the atom or residue. High shade indicates more exposure to the solvent (Fig. S8). The measured binding affinity of compound 1G and the inhibitor of ESR1 receptor in terms of the LibDock score was further re-calculated through different other scoring functions so that to avoid false positive predictions. The calculated docking scores of compound 1G against ESR1 were 99.46, 6.83, 629 and 157.7 for PLP1, Jain, Ludi and PMF scoring functions, respectively. However, for control drug/inhibitor of ESR1, the calculated docking scores were 92.85, 6.02, 674 and 135.86 for PLP1, Jain, Ludi and PMF scoring functions, respectively. These results indicate that compound 1G docking scores were higher than the control drug/inhibitor (Table 3). These results may provide a molecular level of the foundation, to understand the possible mode of action of top hit compound 1G. This infers that compound 1G may be a potential inhibitor of ESR1.
Compliance with in silico pharmacokinetics parameters. The use of in silico methods to predict various pharmacokinetics parameters such as ADMET is intended as a first step to analyze the novel chemical entity, to prevent wasting time on lead candidates that would be toxic or metabolized by the body into an inactive form and unable to cross membranes. In the studied work, different physicochemical properties of the candidate compound 1G were calculated and then compared with that of standard drug topotecan. The calculated molecular diffusion coefficient in water for compound 1G was 0.58, and octanol-water distribution coefficient (LogP/D) was 3.45. The compound 1G was found slightly lipophilic in nature and showed good solubility, but slightly lower than standard anticancer drug topotecan. Results of compound 1G showed a tendency to supersaturate in water, with  These results indicate high intrinsic passive uptake capacity of the liver, which considered good in the sense of pharmacology studies. The calculated volume of distribution of compound 1G was 5.33 L/kg. The compound 1G showed the ability to cross the BBB partition. The brain/blood partition coefficient was found (in logarithm) −0.36, whereas the percent unbound to blood plasma proteins was 4.97. The predicted blood to plasma concentration ratio was 0.72 for Compound 1G. The major CYPs involved were CYP3A4, CYP1A2, CYP2C9, and CYP2C19, and the compound 1G showed potency to inhibit these CYPs. The compound 1G was predicted to act as a substrate of CYP3A4, and the identified sites of metabolism on compound 1G were C13, C16, C21, and C24. The compound 1G was found to be the inhibitor of the CYP3A4-mediated metabolism of midazolam and testosterone. The calculated inhibition constant (K i ) valuations for midazolam and testosterone inhibition were 0.769 and 69.033 µM, respectively. The affinity of compound 1G for CYP450 enzymes in quantitative terms was also evaluated. The calculated K m and V max values provide the knowledge of metabolic rate. The calculated kinetics Michaelis-Menten K m constant for predicted sites of CYP3A4 mediated metabolism was 1.89E + 01 µM, whereas the calculated V max constant for predicted sites of enzyme CYP3A4 mediated metabolism was 1.40E + 01 nM/min/nM. The calculated intrinsic clearance constant (CL int ) for predicted sites of CYP3A4 mediated metabolism was 8.25E + 01 µL/min/mg. The enzyme kinetics Michaelis-Menten K m constant value for CYP3A4 mediated metabolism (human liver microsomes) was 2.36E + 02 µM, while V max constant for predicted sites of CYP3A4 mediated metabolism (human liver microsomes) was 6.70E-01 nM/min/nM. The intrinsic clearance constant for predicting sites of CYP3A4 mediated metabolism (human liver microsomes) was 2.84E + 00 µL/min/mg. These metabolisms related enzyme kinetics data may further be used to calculate the hepatic clearance and in vitro/in vivo relationship. The overall calculated intrinsic clearance in human liver microsomes was 1.82E +01 and 1.64E +01 μL/min/ mg for compound 1G and standard anticancer drug topotecan, respectively. These results suggest that the renal clearance may decrease for compound 1G, but metabolic clearance may increase. Through this metabolic rate, a precise knowledge of elimination rate may be optimized and later can be used to calculate the drug's half-life and total clearance (Tables S13 and S14).
In addition to the CYP450, however, there are a variety of other drug metabolizing enzymes such as oxidases, hydrolases, reductases, and dehydrogenases (oxidoreductases) that can affect the distribution of orally administered compound in the systemic circulation. To study potential metabolism mediated compound interactions in terms of their metabolites, have important implications for both drug efficacy and safety. For this, compound 1G was analyzed computationally for investigation of possible drug metabolism and their metabolites implications. Due to this, predicted results of potential metabolic sites, metabolites, and type of reactions involved for compound 1G are discussed here in details. Results of CYP450 mediated metabolism of compound 1G showed the probability of six possible metabolites by CYP 3A4 enzyme (Fig. 3). Moreover, results also showed the possible metabolic sites and corresponding metabolites for each biochemical reaction. These metabolic reactions for compound 1G were demethylation, glucuronidation, oxidation, sulfation, glucosidation, phosphorylation, and hydration (Fig. 4). However, no sites were identified for (UGT) uridine 5′-diphospho-glucuronosyl transferases family, which catalyze in Phase II metabolism and has the potential to transform small molecules to water-soluble form. The overall results suggest that compound 1G covers, good drug-like properties.
Molecular interactions of compound 1G with druggable proteins. There are some proteins which are reported to be involved in xenobiotic metabolism, detoxification, BBB penetration and channel inhibition. Results indicate that compound 1G may activate the PXR, which regulates the expression of proteins involved in detoxification. Beside this, compound 1G showed the potential to become a substrate, as well as possesses a tendency to inhibit the human P-glycoprotein transporter (Pgp), which is known to be involved in multiple drug resistance (MDR) and BBB penetration. Results also indicate that compound 1G inhibited by the hepatic organic anion-transporting polypeptide (OATP-1B) transporter, thus there may be a little chance of drug-drug interaction (Table S13). These results support the predicted anticancer activity of compound 1G.  Predicted toxicology of compound 1G. Results of in silico toxicity risk assessment for compound 1G revealed no sign of cardiotoxicity and anemia, similar to standard anticancer drug topotecan. Results of compound 1G showed no hERG (human ether-a-go-go-related gene) potassium channel inhibition in human, similar to standard drug topotecan, which otherwise may cause risk of cardiotoxicity. However, results showed that compound 1G may cause mild nephrotoxicity if used for prolonged or at high doses. Drug-induced liver injury (DILI) studies indicate that compound 1G may elevate the level of GGT, SGOT, SGPT, ALP and thus may cause liver necrosis, Cholestasis and may damage the bile duct, On the other hand, level of LDH enzymes thus not cause any risk of ischemic hepatitis. Likewise, results of compound 1G showed predicted rat maximum tolerated dose in feed state was 0.16 g/kg body weight, similar to standard drug topotecan i.e., 0.15 g/kg body weight. The compound 1G predicted to be non-sensitized against toxicity risk parameter allergenic respiratory sensitization in the rat, similar to topotecan. Beside this, compound 1G showed moderate ocular irritancy, mild skin irritant, and weak skin sensitization, in contrary to topotecan, which showed none irritancy, except severe ocular irritancy. However, compound 1G was detected non-toxic for estrogen receptor toxicity in the rat, while topotecan showed a toxic response. On the contrary, compound 1G showed toxic response against androgen receptor toxicity in the rat, similar to topotecan. However, both 1G and topotecan cause non-toxic response against parameter causing phospholipidosis. The compound 1G showed non-mutagenic (Ames) response, similar to topotecan. In terms of pure compound and metabolites, the predicted mutagenicity responses varied against different strains of S. typhimurium ( Table 4). The compound 1G may cause developmental toxicity, similar to topotecan. No carcinogenicity potential predicted for both female and male species of rat and mice. The carcinogenic potency for compound 1G in terms of predicted tumorigenic dose, 50% (TD 50 ) for the mouse and rat were 4.33 and 2.60 mg/kg body weight/day, respectively, similar to topotecan (Table 4). These toxicity risk assessment results overall indicate that compound 1G showed under limit toxicity range, except few parameters, which can be optimized by dose limit experiments. These results will be helpful in setting dose ranges for in vivo small animal's assays.

Enrichment analysis through Systems Pharmacology.
To find out the processes upregulated or downregulated by compound 1G, function-based pathways, compounds association analysis, GO terms were studied in detail, through biological processes, molecular functions, and genes set enrichment scores. This study resulted in a relationship between the effect of compound 1G and biological processes. These EA results for compound 1G are represented through the GO process and key network object (Fig. S9), metabolic networks (Fig. S10), and biological process networks (Fig. S11) and corresponding biological pathway maps. The significance of EA score was represented by p-value (negative logarithm; -log value). Higher p-value represents higher relevance of the entity. Thus, each process ranked based on quantitative p-value and so summarizes the pharmacological and toxic effect at systems pharmacology level.

GO Processes & Key Network Objects.
The key network objects and GO processes were identified for compound 1G through the MetaDrugTM database tool. A total of 27 networks was predicted, however, the top networks were selected based on g-score. Highly positive g-score means, the network is highly saturated with genes ( Table 5). The key networks identified are glutathione s-transferase (GSTA1), sulfotransferase (SULT1A1), huntingtin interacting protein (HYEP), N-acetyltransferase (NAT-1), aryl hydrocarbon receptor nuclear translocator (ARNT2) and the major GO Processes identified was responding to xenobiotic stimulus & xenobiotic metabolic process. A canonical pathway for the top scored network (based on a number of pathways) from active experiments are shown in Fig. S12.
Identified metabolic networks for compound 1G. The drug-induced metabolic changes in the metabolic networks are identified and score in the term of −log (p-value) score. Through this score, a subset of metabolic networks was identified and ranked in Fig. S10, which seems more influenced by the compound 1G. These changes were due to the interactions of some regulatory proteins with the compound 1G or its metabolites. The result indicates that the compound 1G more interferes with the Estradiol & Estrone metabolism. The details interfere are shown here with the pathway details.
The PXR/RXRα also induces the bile acid conjugation enzymes, SULT2A1. This activation of FXR reduces hepatic fat accumulation and the level of plasma triglyceride (Fig. S13).
Bile acid regulation of lipid metabolism. The analysis result, direct that the intracellular bile acid bound with FXR and activate it. On the other hand, bile acids go for reaction with CYP7A1 and make a product of bile acid CoA and later into mono-conjugated bile acid. This product has further transformed into di-anionic bile acids. The targeted FXR, transcriptionally regulate UGT2B4, peroxisome proliferator-activated receptor alpha (PPARα), and SULT2A1. This SULT2A1 transcriptionally regulate the mono-conjugated bile acid. Simultaneously the FXR transcriptionally regulates the PXR, a nuclear receptor which makes a complex subunit PXR/RXRα. The complex PXR/RXRα transcriptionally regulates the cytoplasmic enzyme CYP3A4 and CYP2B6. The CYP3A4 activates the SULT2A1 which transcriptionally regulate the mono-conjugated bile acid and catalyzes a reaction which transforms the mono-conjugated bile acid into di-anionic bile acids. The transcriptionally activated PPARα considered to be the important regulator of intra and extracellular lipid metabolism. The activated FXR increase the lipid oxidation. The FXR targeted the PPARα. The FXR/RXRα transcriptionally activate the gastrotropin, this protein can bind bile acid and play roles in fatty acid uptake, transport, and metabolism (Fig. S14). These results indicate that there was modulation of bile acid receptors by compound 1G, which seems functionally active with a role in lipid metabolism, therefore there is little or no risk of drug-induced toxicity on lipid metabolism.
Neurophysiological process analysis (Circadian rhythm). Screening results showed that compound 1G may also affect the circadian rhythm. To demonstrate the affected neurophysiological processes by compound 1G and its metabolites, circadian rhythm analysis was studied and highlighted the molecular insight mechanism of action. The result of this regulation was analyzed and presented through a canonical pathway. The result showed that the initial regulation was by retinoic acid -related orphan receptor (RORα) and reverse-erythroblastosis (Rev-ERBα) which modulate the neuronal pas domain protein (NPAS2). This NPAS2 also regulated by NAD (+). Further, this NPAS2 bind with cryptochrome (CRY2), which transcriptionally regulates brain and muscle  Table 5. The key network objects and GO processes along with total nodes.   Identification of Drug target networks for compound 1G. The possible mechanism of action of compound 1G was derived by mining the available omics data and thus hypothetically identified the pathway and associated functional genes involved. Results of in silico signal transduction studies suggest that compound 1G or its metabolites may interfere cancer cell signaling process through interaction with transporter FGF4 or its cell membrane receptor FGFR1. This interference resulted in induction of transcriptional co-activation enzyme CBP (CREB-binding protein) in the cytoplasm. The CBP enzyme may transduce this signal through three nuclear membrane receptors namely, ESR1, RAR-alpha/RXR-alpha, and GCR-alpha. The ESR1 predicted to regulate nuclear proteins, namely, COMT, IGF-1, and vascular endothelial growth factor receptors (VEGFR-2). For therapeutic activity, ESR1 receptor transcriptionally activates the Insulin-like growth factor 1 (IGF-1), while inhibiting the COMT. In another signaling pathway, CBP may bind to RXR-alpha and transcriptionally activate the aldehyde dehydrogenase 2 family (ALDH2), thus resulted in inhibition of cell proliferation. The other possible target suggested to be GCR-alpha, which transcriptionally inhibit the IL-4, IL-5, and regulate the macrophage colony-stimulating factor (M-CSF) a nuclear receptor, thus resulted in inhibition of cell division (Fig. 7).
Off-target prediction for compound 1G. Results suggest that compound 1G may interact and interfere with epidermal growth factor (EGF) which well along bind and activates the cell membrane EGFR receptor, and later activate the nuclear C-Jun and SP1 transcription factors. In response to the activation of these nuclear transcription factors, results suggest that it may cause inhibition of cytoplasmic off-target enzyme COMT (which is a known drug target for Parkinson's disease (PD), Alzheimer disease (AD), and Schizophrenia (Fig. 8). These off-target results suggest the future scope of compound 1G for the evaluation of anti-PD, anti-AD, and antipsychotic (against Schizophrenia) activity, which is a subject of further research work.

Conclusion
The quest for high target potency should not be pursued blindly, without an understanding of its relevance to efficacy and efficiency. The strategy used in this study may provide understanding in designing novel and promiscuous Garcinia caged xanthones as anticancer agents. Differentiating and describing the role of important chemical descriptors identified through QSAR modeling gives an idea of key descriptors responsible for the in vitro anticancer/cytotoxic activity. In silico approaches were used to virtually screen top hit compound 1G, and later validated by evaluation through oral bioavailability parameters, ADMET risk screening, docking, in silico pharmacokinetics/pharmacodynamics (PK/PD) screening study, and lastly with systems pharmacology approach. Structure-guided insights of molecular interactions were explored to highlight the multi-level cell networks for biological processes, networks, and signal transduction pathways for Compound 1G. Also highlighted the on and off-targets of Compound 1G. Based on signaling pathways and molecular docking energy, potential targets of compound 1G were suggested. These studied methods can be used as a template work, ahead of smart drug discovery path. By using these QSAR models and through standardization of compounds and centering and scaling of descriptors a set of virtually designed compounds can be predicted as the promiscuous cancer inhibitors. The mechanism of the compound is also describe.