2D QSAR and Atom based 3D QSAR study of Tropomyosin Receptor Kinases inhibition by pyrazol derivatives

 

Arpan Adhikary, Ronak Nair, Lakshya Moukthika, Ruchi Verma*

Department of Pharmaceutical Chemistry, Manipal College of Pharmaceutical Sciences,

Manipal Academy of Higher Education, Manipal-576104, Karnataka, India.

 

ABSTRACT:

Quantitative Structure Activity Relationship (QSAR) studies are tools mostly used in many research areas, including drug discovery process. The tropomyosin receptor kinase (TRK) family are emerging as an important target for cancer therapeutics. The atom based 3D QSAR model and 2D QSAR model were designed and suitable models were generated useful for predicting the tetrahydropyrrolo[3,4-c]pyrazol derivatives prior to their synthesis, developed for predicting the anti-cancer activity against TRKs . The given study indicates the credibility of derived QSAR model by the determination of suitable statistical parameters as we have observed high relationship between experimental and predicted activity values showing ligand molecule larotrectinib with various possibilities of structural modifications to develop potential molecules with significant TRKs inhibitory activity and also predict the activity of any unknown derivative. The data reported by the above QSAR models provides necessary directions for the designing of new TRKs inhibitors against cancer.

 

 

 

1. INTRODUCTION: 

Our bodies are made up of countless number of cells, grouped to form tissues and it is our genes which direct our cells to show growth, division and ultimately causing cell death and thus, our body function normally. However, when there is a change in our DNA, a gene can mutate¹. Mutated genes don’t work properly because the instructions in their DNA are scrambled which can cause the cells that should be resting, to divide and grow out of control which can lead to cancer.This failure of apoptosis (programmed cell death) has the potential to invade or spread to other parts of the body². In this study, we focus on TRKs (Tropomyosin Receptor Kinases) subtype A, primarily a trans-membrane receptor protein (PDB ID: 5KVT) that regulates cell differentiation, proliferation and survival.  The family of TRKs are advancing as major target for cancer therapeutics ³. TRK stimulation leads to activation of a number of intracellular signaling cascades including, the RAS and the phosphatidylinositol-3 kinase (PI-3K) cascades which are cancer causing carcinogenic factor.

 

Larotrectinib is the expected molecule that lines up with the features of Tetrahydropyrolo derivatives in docking study and QSAR analysis is done to forecast their activities⁴.

 

Quantitative Structure Activity Relationship (QSAR) models are models⁵ that can be used to predict the physicochemical, biological and environmental properties of compounds from the knowledge of their structure⁶. QSAR quantifies the relationship between structure and activity on the basis of their physiochemical property⁷. In a QSAR we use descriptors that are eventually used, to convert chemical structures into mathematical variables; statistical methods to derive the relationships between the observations and the descriptors and the quality of the observed data⁸. In this article, we’ve build 3D and 2D QSAR models using TRKs inhibitors data for predicting the activity.

 

2.    METHODOLOGY:

2.1 Computational details:

PHASE 4.3 of Maestro 11.4 version software package of Schrodinger Inc., was used to generate 2D and 3D-QSAR models for tetrahydropyrrolo[3,4-c]pyrazol derivatives as inhibitors of TRKs. A dataset consisting of 42 derivatives were chosen from the mentioned Literature¹⁰ and were used in this study, consisting of minimal concentration of the compounds (IC50) which were then converted to PIC50 by applying the respective formula (PIC50= Log [1/IC50]) which are represented in. The structures were drawn with the help of 2D sketcher option in Schrodinger software.

 

2.2 Ligand preparation and molecular alignment:

The selected compounds were prepared by using LigPrep tool with force field- OPLS_2005 assigned at pH state 7.4  and then ligand alignment was carried out with respective alignment tool ¹⁰ provided by the maestro software. The structures were aligned in such a way that they superimpose one other which helps in observing variations in them. It is a key step in order to make a precise and accurate 3D QSAR model ¹¹.

 

2.3 Datasets:

From the total 42 molecules, 26 of these were selected for training set by keeping its division 75% with PLS factor as five and 7 were selected as test set based on the rule that the activity of the training dataset should range from the most potent to the least potent one ¹². PHASE provides a set of built-in pharmacophoric features such as hydrophobic group, negatively ionizable, positively ionizable, Electron withdrawing and hydrogen bond donor presented in-. The model was prepared using these features to generate sites for all the compounds which can be used to create partial least-squares (PLS) factors of 3D-QSAR model ¹³.

 

2.3 PLS (Partial Least Square) analysis and validation:

The predictive value of the models was evaluated by leave one- out (LOO) to generate cross validated-  r²_cv value(regression), SD(Standard deviation), F( variance ratio), P(significance level of F), RMSE (root mean square error),Q²value (predicted activity) and Pearson-R (relation between the predicted and observed activity for the test set) ¹⁴. 3D-QSAR models are accepted if they fulfill the conditions - r²_cv>0.5 and Q²> 0.6

 

2.4 External validation:

A Q² is mostly useful but not satisfactory factor for the validation of model. In several cases it was observed that model with good values of r²_cv and r² were found out to be disappointing ¹⁵. Although a model gives good prediction on the basis of test set data, it’s still not assured that it will always predict fine new data set of compounds. Thus, an external test with r²_pred (predictive r²) is suggested for prediction.

 

Table 1: Molecules with structures for QSAR study with their measured biological activity, predicted activity and dataset.

Compound no.	Structures	IC50 value (µM)	PIC50 value (µM)	Datasets
17a		1.24	5.9066	Training
17b		1.10	5.9586	Training
17c		1.12	5.9508	Training
17d		0.98	6.0088	Training
17e		1.36	5.8665	Training
17f		0.94	6.0269	Test
17g		0.67	6.1739	_
17h		0.47	6.3279	Training
17i		0.60	6.2218	Training
17j		0.45	6.3468	Test
17k		1.18	5.9281	_
17l		0.19	6.7212	Training
17m		0.20	6.6990	_
17n		0.48	6.3188	Test
17o		0.68	6.1675	_
17p		2.32	6.6345	_
17q		0.52	6.2840	Training
17r		0.94	6.0269	Test
17s		0.88	6.0555	Training
17t		16.36	4.7862	_
17u		1.00	6.0000	Training
17v		6.96	5.1574	_
19a		0.46	6.3372	Training
19b		0.44	6.3565	Training
19c		0.18	6.7447	Training
19d		0.21	6.6778	-
19e		0.35	6.4559	Training
19f		0.95	6.0223	Training
19g		0.71	6.1487	Training
19h		0.029	7.5376	Training
19i		0.014	7.8539	Training
19j		0.069	7.1612	Training
19k		0.096	7.0177	Test
19l		0.017	7.7696	Training
19m		0.017	7.7696	Training
19n		0.095	7.0223	_
19o		0.27	6.5686	Training
19p		0.12	6.9208	Training
19q		0.42	6.3768	Test
19r		0.27	6.5686	Test
19s		0.21	6.6778	Training
19t		0.15	6.8239	Training
Larotrectinib		0.0034	8.4685	Training

 

3. RESULTS AND DISCUSSION:

Atom Based 3D QSAR analysis:

The atom-based 3D QSAR model was developed on the basis of their biological activity and their relation to its structural attributes. The graphical representation of the relation between activity and predicted activity of test and training set is shown in. While creating this model we employed a dataset of 42 molecules for which ligand preparation was performed with OPLS_2005 force field at pH 7.4. Then 75% division was kept for the training set with OPLS factor as five ¹⁶. Scatter plot was constantly checked for best fit line molecules and those molecules which had shown deviation were left out, by which a  test and a training set was built and after completion of model development, the left out molecules were then inserted into the model for predicting its activity. The scatter plot for both test and training set are shown in Figure . The above model was selected due to its high Q² (0.8296) and appropriate r² (0.9797).The atom based QSAR statistics for the selected model with 5 PLS factors are shown at.Due to its high cross validation coefficients this model tends to be precise and efficient.

 

 

 

3D QSAR visualization:

The 3D QSAR study was visualised in fields of Hydrophobic interactions, Negative Ionic interactions, Positive ionic interactions, electron withdrawing) and other factors such that the dark blue(good activity) and red colour(bad activity) regions indicates the positive and negative thresholds on the activity. The visualisations of the QSAR model are shown in 2 ¹⁷. This visualisation gives us the insight about the structural properties present in the ligand molecule and their relation to its biological activity with also the effect of functional groups substitution in the molecule.

 

Table 2. Atom based (QSAR) statistics with five best common PLS factors

Factors	SD	r2	r2cv	r2pred	F	P	RMSE	Q2	Pearson-R
1	0.2926	0.8357	0.6698	0.2359	127.2	2.68e-11	0.16	0.7512	0.9259
2	0.1989	0.9271	0.7898	0.4190	152.7	2.24e-14	0.15	0.7886	0.9191
3	0.1592	0.9552	0.8192	0.5747	163.6	1.18e-15	0.16	0.7506	0.8870
4	0.1392	0.9673	0.8163	0.6536	162.6	5.35e-16	0.15	0.7850	0.9029
5	0.1122	0.9797	0.8090	0.7091	202.9	4.94e-17	0.13	0.8296	0.9160

 

 

Dataset plot for experimental

 

Test dataset plot

 

 

Training dataset plot  data versus predicted data

Figure 1: Scatter plot for Atom Based 3D QSAR showing relation between activity observed and the predicted activity

Table 3: QSAR results of pharmacophoric features

Factors	H-bond donor	Hydrophobic interaction	Negative ionic	Positive ionic	Electron-withdrawing
1	0.032	0.718	0.003	0.023	0.224
2	0.028	0.715	0.003	0.033	0.221
3	0.031	0.682	0.007	0.041	0.239
4	0.028	0.678	0.010	0.039	0.245
5	0.028	0.723	0.008	0.038	0.203

3.1 Hydrophobic/ non-polar visualization:

 

 

(19i)

3.2 Electron withdrawing interaction visualization:

 

 

(19i)

 

 

(Larotrectinib)

 

 

(17e)

3. 3 Hydrogen bond donor interaction:

 

(Larotrectinib)

 

 

(17e)

Figure 2. Atom Based 3D QSAR visualization images

 

2D QSAR analysis and visualization:

Here, we used 2D Canvas which provides seven fingerprint models which are as follows: atom pair, atom triplet, dendritic, molprint, radial, linear and topological models ¹⁸. It is associated with Kernel based PLS (KPLS) in order to generate a precise QSAR model that gives information about the overall favorable and unfavorable attributes of the given dataset. The same set of test and training used in atom-based 3D QSAR was utilized here presented in.

 

Among all models, the linear model was found out to be most suitable which helps in predicting the activity more precisely with the training validating set- r² and standard deviation (SD) values as 0.9858 and 0.0919 respectively and test validating set Q2 and RMSE values as 0.9087 and 0.09906 respectively with 5 KPLS factors. The statistical data of the Fingerprint based 2D QSAR model are presented at ¹⁹. The 2D QSAR fingerprint visualization of the linear model can be seen in. It is shown that the red and blue color regions represent the positive and negative activity effects. The scatter plots for 2D QSAR is shown in .

 

Table 4: Data statistics of the 2D QSAR model

	Training		Test
KPLS factors	SD	r2	RMSE	Q2
1	0.3913	0.6902	0.1596	0.7629
2	0.2201	0.9061	0.07249	0.9427
3	0.1665	0.9486	0.1143	0.8784
4	0.1207	0.9742	0.1061	0.8952
5	0.0919	0.9858	0.09906	0.9087

 

 

Scatter plot for both training and test set

   

Training set plot                                                                                          Test set plot

Figure 3. Scatter plots for Fingerprint Based 2D QSAR

 

       

(Larotrectinib)                                                                               (19i)

Figure 4: Linear fingerprint 2D QSAR model visualization

 

4 CONCLUSION:

The atom based 3D QSAR model and 2D QSAR model were designed and suitable models were generated useful for predicting the tetrahydropyrrolo[3,4-c]pyrazol derivatives prior to their synthesis, developed for predicting the anti-cancer activityagainst TRKs . The given study indicates the credibility of derived QSAR model by the determination of suitable statistical parameters as we have observed high relationship between experimental and predicted activity values showing ligand molecule larotrectinib with various possibilities of structural modifications to develop potential molecules with significant TRKs inhibitory activity and also predict the activity of any unknown derivative. The data reported by the above QSAR models provides necessary directions for the designing of new TRKs inhibitors against cancer.

 

5. CONFLICT OF INTEREST STATEMENT:

The authors declared no conflict of interest.

 

6. ACKNOWLEDGEMENTS:

Authors would like to acknowledge Manipal College of Pharmaceutical Sciences for the facilities to carry out the research work.

 

 

7. REFERENCES

1.      Zhao M, Wang L, Zheng L, Zhang M, Qiu C, Zhang Y, Donshu D, Bing. 2D-QSAR and 3D-QSAR Analyses for EGFR Inhibitors. Biomed Res Int. 2017:4649191 https://doi.org/10.1155/2017/4649191.

2.      Verma R, Bairy I, Tiwari M, Bhat GV, Shenoy GG. In silico studies, synthesis and anticancer activity of novel diphenyl ether-based pyridine derivatives. Mol Divers. 2019;23(3):541–554: https://doi.org/10.1007/s11030-018-9889-1

3.      N. Noolvi M, M. Patel H, Bhardwaj V. A Comparative QSAR Analysis of Quinazoline Analogues as Tyrosine Kinase (erbB-2) Inhibitors. Med Chem. 2011;7(3):200–212. https://doi.org/10.2174/157340611795564213

4.      Goudzal A, Hadaji E, Bouachrine M, El Hamdani H, Ouammou A. QSAR and docking molecular models to predict anti-cancer activity on a series of azacalix [2] arene [2] pyrimidine derivatives as anticancer agents. Mater Today Proc. 2020;51(2).https://doi.org/10.1016/j.matpr.2020.08.002

5.      Ul-Haq Z, Effendi JS, Ashraf S, Bkhaitan MM. Atom and receptor based 3D QSAR models for generating new conformations from pyrazolopyrimidine as IL-2 inducible tyrosine kinase inhibitors. J Mol Graph Model. 2017;74:379–395. https://doi.org/10.1016/j.jmgm.2017.03.016

6.      Pourbasheer E, Amanlou M. 3D-QSAR analysis of anti-cancer agents by CoMFA and CoMSIA. Med Chem Res. 2014;23(2):800–809. https://doi.org/10.1007/s00044-013-0676-3

7.      Kasmi R, Hadaji E, Chedadi O, El Aissouq A, Bouachrine M, Ouammou A. 2D-QSAR and docking study of a series of coumarin derivatives as inhibitors of CDK (anticancer activity) with an application of the molecular docking method. Heliyon. 2020;6(8):e04514.  https://doi.org/10.1016/j.heliyon.2020.e04514

8.      Singh KD, Karthikeyan M, Kirubakaran P, Nagamani S. Pharmacophore filtering and 3D-QSAR in the discovery of new JAK2 inhibitors. J Mol Graph Model [Internet]. 2011;30:186–197. http://dx.doi.org/10.1016/j.jmgm.2011.07.004

9.      Wu T, Zhang C, Lv R, Qin Q, Liu N, Yin W, et al. Design, synthesis, biological evaluation and pharmacophore model analysis of novel tetrahydropyrrolo[3,4-c]pyrazol derivatives as potential TRKs inhibitors. Eur J Med Chem. 2021;223:113627. https://doi.org/10.1016/j.ejmech.2021.113627

10.   Ma J, Zhang H, Zhang X, Lei M. 3D-QSAR studies of D3R antagonists and 5-HT1AR agonists. J Mol Graph Model [Internet]. 2019;86:132–141. https://doi.org/10.1016/j.jmgm.2018.10.013

11.   Wang M, Li W, Wang Y, Song Y, Wang J, Cheng M. In silico insight into voltage-gated sodium channel 1.7 inhibition for anti-pain drug discovery. J Mol Graph Model. 2018;84:18–28.  https://doi.org/10.1016/j.jmgm.2018.05.006

12.   Sadeghian-Rizi S, Sakhteman A, Hassanzadeh F. A quantitative structure-activity relationship (QSAR) study of some diaryl urea derivatives of B-RAF inhibitors. Res Pharm Sci. 2016;11(6):445–453. https://doi.org/10.4103/1735-5362.194869

13.   Dermeche K, Tchouar N, Belaidi S, Salah T. Qualitative Structure-activity relationships and 2D-QSAR modeling of TNF-α inhibition by thalidomide derivatives. J Bionanoscience. 2015;9(5):395–400. http://doi.org/10.1166/jbns.2015.1320

14.   Kristam R, Parmar V, Viswanadhan VN. 3D-QSAR analysis of TRPV1 inhibitors reveals a pharmacophore applicable to diverse scaffolds and clinical candidates. J Mol Graph Model. 2013;45:157–172. http://dx.doi.org/10.1016/j.jmgm.2013.08.014

15.   Zhao Y, Gu W, Li Y. Molecular design of 1,3,5,7-TetraCN derivatives with reduced bioconcentration using 3D-QSAR modeling, full factorial design, and molecular docking. J Mol Graph Model. 2018;84:197–214. http://doi: 10.1016/j.jmgm.2018.07.006

16.   Jiménez Villalobos TP, Gaitán Ibarra R, Montalvo Acosta JJ. 2D, 3D-QSAR and molecular docking of 4(1H)-quinolones analogues with antimalarial activities. J Mol Graph Model [Internet]. 2013;46:105–124. http://dx.doi.org/10.1016/j.jmgm.2013.10.002

17.   Poojita K, Fathima F, Ray R, Kumar L, Verma R. 3D Atom based QSAR model of DprE1 inhibitors as Anti-tubercular Agents. Res J Pharm Technol. 2021;14(11):5903–10. https://doi.org/10.52711/0974-360X.2021.01026

18.   Amata E, Marrazzo A, Dichiara M, Modica MN, Salerno L, Prezzavento O, et al. Comprehensive data on a 2D-QSAR model for Heme Oxygenase isoform 1 inhibitors. Data Br. 2017;15:281–299. https://doi.org10.1016/j.dib.2017.09.036

19.   Poojita K, Fathima F, Ray R, Kumar L, Verma R. Atom based 3D QSAR and Fingerprint based 2D QSAR of Novel Molecules as MmpL3 receptor inhibitors for Mycobacterium tuberculosis. Res J Pharm Technol. 2021;14(12):6321–6329. https://doi.org/10.52711/0974-360X.2021.01093

 

 

 

 

Accepted on 05.08.2023           © RJPT All right reserved

Research J. Pharm. and Tech 2023; 16(10):4681-4690.