Computational Approaches for the Designing of Novel Anticancer Compounds Based on Pyrazolo[3,4-d]pyrimidine Derivatives as TRAP1 Inhibitor

Tumor necrosis factor (TNF) receptor associated protein 1 (TRAP1), a mitochondrial paralog of Heat Shock Protein (Hsp90), is associated with tumorigenesis promotion in different cancers through the maintenance of integrity of mitochondria, reprogramming cellular metabolism and reducing the production of reactive oxygen species (ROS). Therefore, both TRAP1 and Hsp90 are found to be of interest as targeted in the development of cancer therapeutics. In the current research, various computational approaches have been used in the development of TRAP1 inhibitors as anticancer compounds of Pyrazolo[3,4-d]pyrimidine derivatives. The various studies including development of pharmacophore, docking, 3-D QSAR, virtual screening and other studies were performed on 34 different Pyrazolo[3,4-d]pyrimidine derivatives to record the potential ability of these compounds. The required key features for the study is being provided by the pharmacophore study which provide DHHRR_1 hypothesis. 3D QSAR (atom-based analysis) was performed with different 34 pyrazole derivatives, which has been divided into two groups i.e. training set and test sets, provide the knowledge of the involvement of various elds of atom-based QSAR. The statistically signicant model for this QSAR model was determined by partial least squares regression (PLS) method for which R 2 =0.96 and Q 2 = 0.57 would be considered signicant. The LOO cross-validation R2 CV = 0.58 was used for the validation of QSAR model. Based on the virtual screening study protocol for the optimized binding interaction with TRAP1 kinase receptors (PDB ID: 5Y3N), the compounds ZINC05297837, ZINC05434822 and ZINC72286418 were produced. The maximum XP docking scores (-11.265, -10.532, -10.422, -10.827, -10.753) were observed for potent pyrazole analogues (42, 46, 49, 56, 43) by docking study representing their possible signicant interactions with amino acid residues (ASP 594, CYS 532, PHE 583, SER 536). distribution, metabolism, and excretion (ADME) analysis was carried out providing the key information for the newly designed compounds and their drug ability. The results of the docking study were correlated with the 3-D QSAR analysis revealed the active conformation of TRAP1 inhibitors which is helpful and important for activity performance with future perspective. enumeration module Schrödinger implemented for R-group based enumeration of the Pyrazolo scaffold. Drug-Like lters like, REOS and PAIN’s series were used for separating the compound having reactive functional groups. The obtained drug-like compounds were further processed for preparation of ligand along with the minimum energy with the help of OPLS3e force eld. Additionally, the docking of nal screened compounds was performed in TRAP1 crystal structure in ligand-binding cavity through Glide SP protocol, resulting docking poses. From these different docking poses, best 50 poses were selected from different enumeration for further XP docking protocol providing the XP descriptors. This help in describing contribution of each atom in terms of penalties and rewards for docking energy. Enrichment calculations were performed for 1000 decoy compounds (from DUD.E database) and 30 Compounds (XP best poses) with the help of Schrodinger software while docking were performed by using XP protocol. The results obtained help in predicting validation of docking protocol as Receiver Operating Characteristic (ROC) curve demonstrate R2= 0.92. results with the 3D-QSAR analysis can get more potential compounds as TRAP1 inhibitors. The enumeration on different positions of pyrazole analogues produced compounds with best docking score may be used as synthesis in research laboratory.


Introduction
TRAP1 is a 90 kDa mitochondrial paralog Hsp90 and bioenergetic regulator which is closely related to the tumorigenesis promotion in a variety of cancers [1,2].TRAP1 helps in the maintenance of mitochondrial integrity, thus smooths the progression of a cell death against cellular stresses which is obtained by reduced ROS production along with reprogramming cellular metabolism.These two factors make cancer cells to get better adaptation in the harsh tumor microenvironments [3][4][5].TRAP1 inactivation encourages cancer cells to undergo substantial apoptosis, in-vitro and in-vivo, hence numerous targeting mitochondrial TRAP1 inhibitors have been developed [6].
TRAP1 is an imperative bioenergetic regulator due its ability to inhibit both cytochrome oxidase and succinate dehydrogenase (SDH) [7][8][9].TRAP1 also have the ability to provide resistance to oxidative stress [9] and counterbalance permeability of mitochondrial transition and consequent cell death.
In the present work, attempts have been made to carry the computational analysis for the different 34 pyrazole analogues, as reported in literature.3-D pharmacophore mapping was used in order to identify the important pharmacophoric features accountable for biological activity.The role of individual atom contributing in the model development was basically based on the atom-based model development.The virtual screening studies provide the information about the potential effects of various ZINC compounds against the TRAP1 which are comparable to dataset.The important molecular interactions with the TRAP1 active site were studies by molecular docking studies for the various surrounded amino acids.The novel effective inhibitors of TRAP1 target was determined by using various computational approaches, like pharmacophore mapping, 3-D QSAR, virtual screening and molecular docking which provide the harmonizing information to support the development of potent inhibitors.test set, respectively.The compounds clustering was carried by PLS factor of 5 [23,24].Table 3 summarized the different parameters of QSAR model which are as Factors: number of different factors for partial least squares regression model, SD: regression standard deviation, R2: value for regression, R2 CV: crossvalidated R2 value calculated though the predictions acquired by a leave-N-out approach, F: Variance ratio, higher statistically signi cant regression are represented by larger F values, P: signi cance level of variance ratio, a greater degree of con dence are represented by smaller values, RMSE: root-meansquare error of the test set, Q2: for the predicted activities of test set, Pearson-R: for the predicted activities of the test set.Atom type fraction segment displayed the fraction due to each atom type in the QSAR model for each number of PLS factors used in the model.The con rmation of least diversity in the biological activities between molecules of training set through the scatter plot obtained by plotting actual activity against predicted activity [25, 26].2.4.2Generation of Contour maps The contour maps help in predicting the favourably or unfavourably interaction of aligned molecules with the receptor for biological activity and correspond to the spatial arrangement of aligned molecules.In Field based model, regions with favourable steric elds are represented by green contours and regions with unfavourable steric elds are represented by yellow contours.Moreover, the blue and red contours highlight the positions where electropositive groups and electronegative groups would be favourable, respectively.Thus, it is clear that biological activity will be greater when there is more steric bulk near green, less steric bulk near yellow, more positive charge near blue and more negative charge near red.Hydrogen bond donor contour map, donor bulk near purple is favourable, but the donor bulk near cyan is unfavourable for greater biological activity.For the hydrogen bond acceptor contour map, acceptor bulk near red is desired and acceptor bulk undesired near magenta for improved biological activity [27].Whereas in Atom based model blue cubes represent increase in activity and red coloured cubes represent decrease in activity by a particular group.The contour maps have been described as follow:

Materials And Methods
The such as squared cross-validation coe cient (q2), squared non-cross-validation coe cient (r2) and predictive r2 (r2pred), standard error of estimate (SEE).The developed model was tested for internal quality which was based on the q2 value with an acceptance criterion of >0.5 statistically for signi cant model.The r2 provide the relative measure of the t using regression equation whose value near to 1.0 illustrate the best t of regression.Standard error of estimate is supportive in conveying the information about the variation of residuals or regression line [28,29].2.6 Virtual screening studies on ZINC database The virtual screening study through ZINC database was performed by using pharmacophore hypothesis DHHRR_1 and most active compound 48 (Supplementary le S1).The ZINC database was used as drug like lters to download the 7543 molecules.The prediction of target was further performed by using Swiss Target Prediction which is a freely accessible tool for receptor database (Supplementary le S2) through which the target exploration become more convenient and useful.Swiss Target Prediction was used to predict the various protein targets among which the TRAP1 was the topmost suitable target for different molecules taken.Thus, the screening of molecules was further performed through molecular docking having standard and extra-precision mode against TRAP1 using GLIDE module of Schrodinger [30].2.7 Docking study Molecular docking studies on pyrazole analogues with TRAP1were carried out by using Glide module software (Schrodinger Maestro v12.1).Protein data bank (PDB ID:5Y3N) was used for determining the protein structure which were further processed through "protein preparation wizard'' (Maestro wizard v12.1).Both, the generating states as well as re nement step were helpful in automatic addition of H atoms along with some important bonds at missing sites of protein molecule.The re nement step is crucial as it involve in the optimization of H bonded groups, dehydration and restrained minimization by using default force eld OPLS_ 3e.The completion of optimization process was followed by processing of receptor grid to calculate binding pocket of receptors.Various docked ligand conformations were observed in docking results showing their binding energy scores.The ranking on the basis of scores were given representing high rank for lesser scoring conformation [31,32].2.8 ADME property predictions ADME properties were determined by using Swiss ADME and Schrodinger ADME online tool helping in ligands selection with drug-like properties.Synthetic accessibility: from 1 (very easy) to 10 (very di cult) [35].Default settings were employed for these calculations.2.9 Enumeration study R-group enumeration module of Schrödinger was implemented for R-group based enumeration of the Pyrazolo scaffold.Drug-Like lters like, REOS and PAIN's series were used for separating the compound having reactive functional groups.The obtained drug-like compounds were further processed for preparation of ligand along with the minimum energy with the help of OPLS3e force eld.Additionally, the docking of nal screened compounds was performed in TRAP1 crystal structure in ligand-binding cavity through Glide SP protocol, resulting docking poses.From these different docking poses, best 50 poses were selected from different enumeration for further XP docking protocol providing the XP descriptors.This help in describing contribution of each atom in terms of penalties and rewards for docking energy.Enrichment calculations were performed for 1000 decoy compounds (from DUD.E database) and 30 Compounds (XP best poses) with the help of Schrodinger software while docking were performed by using XP protocol.The results obtained help in predicting validation of docking protocol as Receiver Operating Characteristic (ROC) curve demonstrate R2= 0.92.

Selection of best pharmacophore hypothesis
All the selected compounds (compounds 1 to 34) from the database were screened to get ve probable common pharmacophore features from the list of variants i.e. 2 aromatic rings, 2 hydrophobic interaction, and 1 hydrogen bond donors.The mentioned features were supposed to have essential role in inhibitory ability of different compounds towards the target.Among the 20 hypotheses generated by PHASE module, DHHRR_1 hypothesis was considered to be best depending on a scoring function mentioned in Table 2.

Pharmacophore model evaluation
The pharmacophore model quality was calculated by using two evaluation tools i.e. percent screen plot & ROC plot.Percent screen plot represents the plot of the percentage of actives recovered and percentage of ligands screened for the hypothesis.ROC plot represents the plot between the true positive rate (sensitivity) and false positive rate (speci city) for various cutoff points.A test is considered to be perfect discrimination when it does not have any overlapping in two distributions and test has a ROC curve representing 100% speci city and 100% sensitivity by passing through the left upper corner.The more closely the position of curves in left upper corner represents the higher overall accuracy of the method.Both, percent screen plot as well as ROC plot was found to be an extreme left corner recommending the better accuracy of the generated hypotheses by PHASE module, as shown in Fig 3 (A

Selection of atom based QSAR model
The QSAR results showed the important statistics of the t for both test and training set.In table, each row shows the hypothesis results.Lines within each row is for regression models having a speci c value for least squares factors and clustering compounds has been performed by PLS factor of 5. Different statistical parameters (SD, R 2 , P, F, Q 2 , RMSE and Pearson-R) in QSAR model was used taken in to account for reliable predictions and evaluation of QSAR model.The value of R 2 is required and high R 2 is essential for a model, but it alone does not provide the su cient condition for ideal QSAR model prediction.
Thus, predictive ability Q 2 values have to be chosen for best QSAR model prediction.Based on these parameters, ve different models were developed by module and shown in Table 3.Among the ve models, fth model was found to be signi cant model owing to higher values of 0.57, 0.96 and 0.58 for Q 2 , R 2 & R 2 CV values, respectively.Though the higher value for SD (0.46) and RMSE (0.64) was recorded but very low values of 0.08, 0.34 and 0.08 for Q 2 , R 2 & R 2 CV values respectively eliminates the probability of rst model selection.The required statics for atom-type fraction are reported in Table 4. Similarly, Table 5 represent the predicted pIC 50 , actual pIC 50 and residual values for generated models.The atom type fractions map provide the information of fractions of each atom of the training set affects the activity and is shown in Fig. 4. The uniform distribution of training set obtained by using scatter plot of displayed module passing through the origin (0, 0) as straight line is shown in Fig. 5.

Contour map analysis
Contour maps help in predicting the biological activity and its correlation with various substituents on the core moiety (Fig. 4) and help in determining the effect of adding substituent' on the biological activity.Increase in biological activity is represented by blue colour while decrease in biological activity is represented by red colour occlusion map.Among the 34 compounds, the most active compound was selected on the basis of high survival value of DHHRR_1 of atom-based 3-D QSAR contour maps.Increase in activity is accountable due to substitution of electron withdrawing group on phenyl ring attached to Pyrazolo [3,4-d]pyrimidine, suggesting that substitution of various groups like -CN, -NO 2 , CF 3 , -NR 3 , -COR -X, etc on phenyl ring is answerable for augmented activity.Further, enhanced anticancer activity could be obtained by addition of hydrogen bond donor group at Pyrazolo [3,4-d]pyrimidine ring.Moreover, the hydrophobic group cover up the larger part of the ring and accountable for mixed activity.

Results of molecular docking
Molecular modelling was performed to examine the possible interactions between protein and most potent derivative through the comparative modelling by using the Schrodinger Glide module.The inhibition of enzyme activity depends on the possible interactions of inhibitors with various amino acid residues of targeted protein of interest.Docking was performed for all analogues to study the binding cavity of TRAP1 (PDB ID: 5Y3N) and are shown in Fig. 6 and Fig. 7.The H-bond are shown by purple colour arrows and Π-Π stacking interactions are shown by purple green colour arrows.The possible bond interactions of compound 42 with amino acid residues PHE 201, GLY162, ASN119, ASP158, PHE205 and TRP231 has been observed in the study.Similarly, the derivative compound 49 (XP docking scores value is -11.353)found to have possible critical interactions with PHE 201, ASN119, ASP158 and PHE205 (Supplementary le S3).Further, the binding interactions of compound 43 has been observed with PHE 201, ASP158, GLY162 and PHE205 while in compound 56 interactions with PHE201, GLY202 and ASP158 amino acids were detected.These interactions were essential for TRAP1 inhibitory activity.

Results of virtual screening
The virtual screening study has been performed by means of pharmacophore hypothesis DHHRR_1 utilizing ZINC database, resulting in the screening of total 2832 compounds with the help of Lipinski's rule of ve.These screened compounds were further used in high throughput virtual Screening (HTVS) docking methodology.The best 20% compounds from HTVS were subjected to SP docking.Similarly, top 20% screened compounds from SP docking were further subjected to XP docking (Supplementary le S4).Total 16 compounds were screened through SP docking in which top hits, namely ZINC05434822, ZINC72286418, ZINC05297837, ZINC59358929 and were found with docking scores -11.97, -10.73, -9.98, -9.88, respectively.These compounds were taken into consideration for the further study, as nal ZINC compounds.These compounds were evaluated in terms of binding interaction energy by MMGBSA.Among these four compounds, ZINC05297837 showed interaction with amino acid residue PHE205, TRP231 and ASN171 via phenyl ring (Fig. 8) while the ZINC05434822 show interaction with PHE 201 and ILE161 in the same cavity as shown by crystal ligand (Fig. 8).The compound ZINC59358929 showed binding interactions with ASN119, PHE201, PHE 205 and TRP231 which is considered signi cant for showing activity while the compound ZINC72286418 showed signi cant interactions with ILE161, PHE201 and PHE205 (Fig. 9).
The binding pocket residues are found to be similar as were obtained from binding of active compounds 42, 46, 48, 49, 56 and crystal ligand.The docking simulation study was further validated by checking the RMSD value which should be less than 2 Å.

Prediction of ADME properties
The ADME properties were determined by using Schrodinger ADME and Swiss ADME tool for obtaining best scoring of dataset and ZINC compounds and shown in Table 7 to 12.All compounds showed signi cant ADME properties, like number of hydrogen bond donor between 0-3, number of hydrogen bond acceptor 7, number of rotatable bond 4-9, like molecular weight were <500, and molar refractivity of about 125 and are considerable (Supplementary le S6 and Supplementary le S7).Lipophilicity pro le of selected compounds represented the lipophilic character along with the high GI absorption, but none of the said compounds possess the ability to cross blood brain barrier representing no toxicity of selected compounds.Compound ZINC72286418 was found to be soluble as determined by solubility pro le of ZINC derived compounds while others were moderately soluble in water.The synthetic convenience of all the compounds was in good range (Supplementary le S5, S6, S7 & S8).

Optimization of novel ligands
The optimization and development of novel TRAP1inhibitors can be performed by using 3-D QSAR and molecular docking studies.Here, results obtained by 3-D QSAR study has been graphically represented as the structure activity relationships (SARs) of Pyrazolo [3,4-d]pyrimidine core with different possible substituent's (Fig. 10)

Analysis of R group enumeration
On the basis of optimized structure several derivatives have been enumerated through R group enumeration study of Schrodinger software.The compound structures have been described in Table 12 with their XP docking scores.These compounds are novel derivatives of Pyrazolo [3,4-d]pyrimidine having good docking scores.

Conclusion
In the present study, development of pharmacophore hypothesis, QSAR, virtual screening and enumeration study has been performed for determining the potential inhibitors against TRAP1.The best hypothesis generated was DHHRR_1, which has been taken for virtual screening study through ZINC database.
The 3D QSAR study determines the best statistical values after many times trial by changing the training and test set molecules.The resultant contour maps determine the features such as electrostatic, hydrogen bond acceptor, hydrogen bond donor and positive ionic, participate in activity.The docking study of potent pyrazole analogues (42,46,49,56,43) showed highest XP docking scores (-11.265,-10.532, -10.422, -10.827, -10.753).Docking study showed the interactions with amino acids such as PHE 583, CYS 532, SER 536, ASP 594 important for activity.The ADME properties showed the important physicochemical properties of the molecules.The virtual screening study performed on ZINC database produced compounds ZINC05434822, ZINC72286418 and ZINC05297837 showed essential binding interaction with receptor TRAP1 (PDB ID: 5Y3N).Correlating the docking results with the 3D-QSAR analysis can get more potential compounds as TRAP1 inhibitors.The enumeration on different positions of pyrazole analogues produced compounds with best docking score may be used as synthesis in research laboratory.

2. 3
Pharmacophore hypothesis generation Relation between the chemical features and structural similarities for mentioned 34 compounds provide the chance of generation of possible 20 hypotheses which explain the binding ability of active molecules with receptors, having box size of 1 Å with 2 Å minimum inter site distance.The features up to 5 were set helping in generation of maximum variants supporting in the establishment of common pharmacophore hypothesis [18].The different parameters for the pharmacophore hypotheses include: (1) Phase hypothesis score: rank-orders a new scoring function hypotheses helpful in providing the knowledge of performance in virtual screening and quality of ligand alignment, provide easy recognition of multiple binding modes by training against diverse known actives through perception of common pharmacophore; (2) Site score: help in providing the intimacy of superimposition of site point to pharmacophore of the structure; (3) Survival score: act as blending terms for the number of matches providing the relation between the relative energy and activity of the reference ligand; (4) Selectivity score: provide negative logarithm of part of molecules in the Index help in matching the hypothesis; (5) Average outranking: active adjusted rank minus one, outranking decoys are calculated for every docked active and then averaged; (6) Receiver operating characteristic (ROC): aid in data analysis as an indicator of model performance providing the demarcation of active site from inactive compounds; (7) Vector score: average cosine of the angles between the analogous pairs of vector features (donors, acceptors and aromatic rings) in different associated structures; (8) Active matched: give knowledge of number of active ligands matching the said hypothesis [19-22].2.4 Model developments by 3-dimensional QSAR study 2.4.1 Atom based QSAR Atom based 3-D QSAR model help in predicting the activities of other molecules.It was developed by Schrodinger Maestro v12.1 from a set of aligned ligands.The compounds have been divided in the ration of 70% and 30% in the training set and atom-based 3-D QSAR model visual representation; (a) electron withdrawing, (b) hydrogen bond donor, (c) hydrophobic, (d) positive ionic where the positive coe cient (increase in activity) is represented as blue coloured cubes while negative coe cient (decrease in activity) is represented as red-coloured cubes represents.Field contour maps; (a) Electrostatic elds: blue as favoured electropositive and red as disfavoured electronegative.(b) Hydrogen bond acceptor eld: red as favoured and magenta as disfavoured.(c) Hydrogen bond donor eld: purple as favoured and cyan as disfavoured.(d) Steric eld: green as favoured and yellow as unfavoured.2.5 3-D QSAR model evaluation The 3-D QSAR model evaluation was carried out by taking key statistical parameters Lipinski (P zer) lter: implemented as MW≤ 500, MLOGP ≤ 4.15, N or O ≤ 10, NH or OH ≤ 5 [43].Ghose lter: implemented as 160 ≤ MW ≤ 480, -0.4 ≤ WLOGP ≤ 5.6, 40 ≤ MR ≤ 130, 20 ≤ atoms ≤ 70 [33-34].Leadlikeness: implemented as 250 ≤ MW ≤350, XLOGP ≤3.5, Number of rotatable bonds ≤ 7; ) & Fig 3 (B).

Table 3 :
Statistical data of QSAR model

Table 5 :
Actual, predicted pIC 50 and residual values of generated models

Table 6 :
Docking scores of active compounds and ZINC screened compounds with their MMGBSA scores S. No. Compound name Log P o/w (XLOGP3)

Table 10 :
Water solubility pro le of ZINC database and other active

Table 11 :
Pharmacokinetics results of ZINC database and other active compounds