Deep Learning Model to Identify Potential Acetylcholinesterase Inhibitors: A Case Study of Isolated Compounds From Pongamia pinnata (L.) Pierre

Abstract

Acetylcholinesterase (AChE) plays an essential role in the cholinergic pathways in Alzheimer's disease. This study used a deep learning model as a powerful virtual tool for discovering AChE inhibitors. The model showed 94.3% accuracy, 97.1% precision, 95.9% recall, and 86.2% specificity. A list of bioactive compounds extracted from Pongamia pinnata (L.) Pierre was selected as the test dataset. Four candidates were selected for in vitro: pongapin, ovalichromene B, gamatin, and pongaglabrone. These flavonoids showed inhibitory effects with half-maximal inhibitory concentration (IC₅₀) values between 19.8 and 63.5 μg/mL. In molecular analyses, these compounds showed noticeable interactions with the AChE catalytic residues Ser203 and His447 and satisfied acceptable drug-like properties and other druglikeness parameters. This study has shown that a deep learning approach can accurately predict potential compounds targeting AChE, and P. pinnata is a promising medical plant for Alzheimer's disease.

Keywords

deep learning acetylcholinesterase Alzheimer's disease molecular docking ADME

Introduction

Alzheimer's disease is a common cause of dementia, and the disease is being described for the first time in the early 20th century by the psychiatrist Alois Alzheimer. Patients progressively lose their memory and ability to focus. In addition, their ability to navigate becomes increasingly complex, making it difficult to perform daily activities. However, the impact of Alzheimer's disease on patients can be reduced by different drugs and non-pharmacological methods.¹ Therefore, developing natural compounds that influence the therapeutic process in Alzheimer's patients has been identified as a potential future research direction.

Acetylcholine (ACh), found in the neural system of many animals and humans, acts as a neurotransmitter. Acetylcholinesterase (AChE) is a cholinergic enzyme that catalyzes the hydrolysis of ACh neurotransmitters into ACh or other choline esters. Several studies have investigated AChE in the tissues related to Alzheimer's disease, showing changes in AChE activity in the brain, cerebrospinal fluid (CSF), and blood of Alzheimer's patients.^2,3 AChE and other enzyme components form a plaque-causing complex in older adults. In addition, the neurotoxicity of amyloid is increased in the presence of AChE.⁴ Therefore, the search for AChE inhibiting compounds has been identified as a potential research direction.

Drug discovery and development is a complex and expensive process. It costs over US$2.6 billion to develop a new drug.⁵ Computational methods have been used to discover new pharmacological candidates through virtual screening analyses.^6‐8 Recent advances in computer science may accelerate and improve pharmaceutical research and development. Deep learning is a form of artificial intelligence (AI) that uses the power of computers and algorithms to solve tasks expediently.⁹ Rapid and remarkable progress has been made on various medical problems through deep learning and large-scale annotated datasets.^10‐12

Medicinal chemistry has adopted various forms of deep learning in drug design with varying degrees of success. For example, many drug features are predicted using the quantitative structure–activity relationship (QSAR) method, such as solubility and bioactivity. Recently, many public datasets have been made available for screening and testing the activity of chemical compounds in drug development and research and are particularly useful for training models to predict the bioavailability of other compounds. The implementation of deep neural networks has resulted in an unprecedented acceleration in this field. This study uses a deep learning approach to classify chemical compounds with bioactivities on AChE.

Pongamia pinnata belongs to the Fabaceae family. Previous studies have found its oil to possess activities useful for treating tumors, skin diseases, abscesses, arthritis, wounds, ulcers, diarrhea, hyperglycemia, and blood lipids.^13‐15 In addition, the toxicity and inhibition of P. pinnata for pests such as Pediculus humanus capitis, Callosobruchus chinensis, and Odototermes obesus have been studied in the laboratory.^16‐19 Importantly, compounds isolated from P. pinnata have shown anti-Alzheimer's activity.²⁰ However, the biological activity of P. pinnata on AChE remains unknown. Therefore, this study uses the deep learning model and AChE inhibition assay to screen potential natural AChE inhibitors isolated from P. pinnata.

Results and Discussion

Deep Learning Model

A dataset of published experimental experiments was retrieved from the ChEMBL database and previous studies.^21,22 Half-maximal inhibitory concentrations (IC₅₀) were converted to μg/mL. The dataset was classified into 2 groups using a cut-off value of 50 μg/mL: active (IC₅₀ < 50 μg/mL) and inactive (IC₅₀ ≥ 50 μg/mL). The division of the dataset into the 2 groups was unbalanced. Therefore, we used resampling for the training dataset to account for the unbalanced groups. We developed a 1D convolutional neural network (CNN) model to predict 2 classes of bioactive compounds inhibiting AChE using the compound's molecular descriptors. The structure of the proposed deep CNNs is presented in Figure 1. The proposed CNN model consists of 6 main convolutional layers, following 6 activation layers and 6 pooling layers. The previous layers are used for feature extraction, after which we added neural and dropout layers (0.3). The activation layers are connected by a rectified linear unit (ReLU) function. Finally, the last layer has 1 node connected to the previous layer by the softmax activation function.

Figure 1.

The structure of the deep convolutional neural model.

The model was trained using 1000 epochs with batch_size = 25. The results showed that the accuracy of the training and validation datasets was above 90% (Figure 2).

Figure 2.

Accuracy and loss of deep learning model (1000 epochs).

It can be seen that after the model was trained for 200 epochs, the accuracy rate remained unchanged. However, while the training loss has remained virtually unchanged, the validation loss increases. Therefore, the model may be overfitting when trained using many epochs. Therefore, we decided to train the model with 200 epochs to optimize its parameters.

The model was trained using 200 epochs with batch_size = 25, and learning_rate = 0.0001. The results showed that the accuracy of the validation dataset is 94.3%, and the validation loss is similar to the training loss (Figure 3). The confusion matrix of the validation dataset is presented in Table 1.

Figure 3.

Accuracy and loss of deep learning model (200 epochs).

Table 1.

The Confusion Matrix of the Validation Dataset.

		Predicted Values
		Active	Inactive
Real Values	Active	847	36
Real Values	Inactive	25	156

Following the confusion matrix, we computed the average values of accuracy, precision, recall, and specificity, defined as follows:

Accuracy = \frac{T P + T N}{T P + T N + F P + F N}; Precision = \frac{T P}{T P + F P}; Recall = \frac{T P}{T P + F N}; Specificity = \frac{T N}{T N + F P}

where TP, TN, FP, and FN represent the number of true positive, true negative, false positive, and false-negative cases, respectively.

The average results showed 94.3% accuracy, 97.1% precision, 95.9% recall, and 86.2% specificity (Table 1), highlighting the model's predictive capability. Furthermore, the screening method identified candidates with effective biological activity that can reduce the time and cost of unsuccessful testing in the experimental research phases. This result indicates that our model favors active compound selection over inactive compound selection.

In this study, our model showed precision and recall values greater than its accuracy values, indicating that the deep learning model can be used for the virtual screening of compounds that inhibit AChE. Therefore, this deep learning model was used to screen natural compounds isolated from P. pinnata; 4 compounds were identified and categorized in the “active” group, which were selected for the in vitro experiments: pongapin (1), ovalichromene B (2), gamatin (3), and pongaglabrone (4) (Figure 4).

Figure 4.

Chemical structure of the 4 compounds.

AChE Assay

The AChE inhibitory activities of the 4 selected compounds were assessed at various concentrations using galanthamine as the positive control. All 4 compounds showed inhibitory activities against AChE with IC₅₀ < 100 μg/mL. Among them, pongaglabrone (4), ovalichromene B (2), and gamatin (3) showed moderate AChE inhibitory activities with IC₅₀ values of 19.8 ± 1.7, 24.4 ± 1.3, and 36.7 ± 4.3 μg/mL, respectively. In contrast, pongapin (1) showed only slight AChE inhibitory activity with an IC₅₀ value of 63.5 ± 4.4 μg/mL. Galanthamine, the positive control, produced a value of 2.0 ± 0.1 μg/mL.

Molecular Docking Analysis

Molecular docking is a critical method for determining the interaction between proteins and drug candidates.^6‐8 We used the AutoDock Vina software to perform molecular docking to determine the structural and activity interactions between AChE and the candidate compounds. The 3D crystal structure of AChE (protein data bank [PDB] ID: 4EY7) was selected as the representative sample for further structure–activity relationship study. The active site of human AChE is located at the bottom of a 20 Å deep, narrow gorge found in different distinct domains.^23,24 The peripheral anionic site is located near (18 Å) the active gorge's site. In this study, the grid box was set to the protein's active site based on the interaction site of Donepezil, the crystal ligand of the protein crystal (PDB ID: 4EY7). First, to demonstrate that the selected protein structure (PDB ID: 4EY7) and docking protocol are reliable, the ligand was re-docked to the target protein in the crystal structure, and the root-mean-square deviation (RMSD) of docking pose against the crystal pose was assessed. The result showed that the docking pose has a similar conformation and orientation as the crystal pose with an RMSD of 0.316 Å (Figure 5). Previous studies have considered an RMSD within 2.0 Å to be suitable and valuable for docking experiments, indicating that this docking protocol is suitable for use in our study.

Figure 5.

Docking (yellow) and crystal (green) poses of the ligand in the crystal structure of acetylcholinesterase (AChE).

The molecular docking method was used to analyze the mechanical interaction of AChE with compounds 1-4, which have a similar structure to flavanone. In previous studies, the amino acids Ser203, His447, and Glu334 were considered a catalytic triad of AChE. We found that the carboxyl group of the 4 compounds plays an essential role in binding energy with AChE. The chemical structure of pyrano flavanones can be seen in ovalichromene B (2) with a pyran group compared to other compounds. Ovalichromene B (2) interacted with AChE at Asn87, Ser125, Tyr337, and Tyr341, with 2 hydrogen bonds and 2 hydrophobic interactions (Figure 6).

Figure 6.

Interactions between ovalichromene B and acetylcholinesterase (AChE).

Compounds 1, 3, and 4 have a structure with 2 furan groups, and the cyclic structure of pongaglabrone (4) consists of carbon and oxygen atoms. The oxygen atom in the furan group was found to interact with AChE through hydrogen bonds. For example, the interaction between AChE and gamatin (3) includes a π-donor hydrogen bond to the key amino residue Ser203 with a distance of 4.61 Å. In addition, among 5 hydrophobic interactions, gamatin (3) also has a π-π interaction with His447 with a distance of 5.64 Å (Figure 7). Moreover, pongaglabrone (4) was found to interact with AChE at the oxygen atom in the cyclic structure of key amino acid residue Ser203 with a distance of 4.61 Å and through 2 π-π interactions at Phe338 and Trp286 (Figure 8). Furthermore, eliminating the methoxy group from the molecular framework minimized its size. Consequently, the molecule should be able to access the active site's deep, narrow gorge more efficiently, explaining why pongaglabrone (4) showed the strongest biological activity among the 4 candidates.

Figure 7.

Interactions between gamatin and acetylcholinesterase (AChE).

Figure 8.

Interactions between pongaglabrone and acetylcholinesterase (AChE).

These interactions are similar to those between pongapin (1) and AChE, which was also found to interact with AChE at Trp286, Tyr341, and Phe297 with 3 hydrogen bonds and at Gln291 and Glu292 with 2hydrophobic interactions (Figure 9). Compared to the other compounds, pongapin (1) had the weakest binding affinity, explaining the correlation between the docking results and experimental AChE inhibition.

Figure 9.

Interactions of pongapin and acetylcholinesterase (AChE).

The interaction analysis showed that the 4 compounds (1-4) form several significant interactions with AChE (Table 2). Moreover, most have a noticeable interaction with the catalytic dyad Ser203 and His447, and essential amino acids Tyr337 and Trp86, critical residues for the enzymatic activity of AChE.

Table 2.

Docking Results of 4 Flavonoids Toward Acetylcholinesterase (AChE) Protein.

No.	Compound	Docking score (kcal/mol)	Hydrogen bond	Hydrophobic interaction
1	Pongapin	−9.2	Gln291, Glu292	Trp286, Tyr341, Phe297
2	Ovalichromene B	−11.2	Asn87, Ser125	Tyr337, Tyr341
3	Gamatin	−11.9	Phe296, Phe295, Gly122, Ser293, Tyr124, Ser203	Val294, His447, Trp286, Tyr341, Trp86
4	Pongaglabrone	−11.7	Phe295, Arg296, Ser203, Tyr124, Tyr341	Phe338, Trp286

Absorption, Distribution, Metabolism, and Excretion Properties

Lipinski's rule of five is a method for determining the druglikeness of active chemicals and determining whether they are orally active or not. The oral bioavailability of drugs with good membrane permeability and hydrophobicity is shown by the Log P, TPSA, MW, HBA, and HBD values, with appropriate nRotB and MR values indicating that medicines are absorbed well in the intestine and have a high oral bioavailability.

The results presented in Table 3 show that the 4 compounds possess acceptable drug-like properties and other druglikeness parameters (MW ≤500 Da, Log P <5, TPSA <140 Å, nHBD ≤5, and nHBA ≤10). These results show that these compounds have the potential to become drugs with good oral bioavailability. In addition, the Log S values are between −4.61 and −4.43, indicating moderate solubility in water, and TPSA values are below 140.²⁵ Consequently, the 4 compounds were predicted to have excellent gastrointestinal absorption. The drug molecule's stereo-specificity is a characteristic of nRotB, which was determined to be under 10.

Table 3.

Drug Properties of Isolated Compounds Analyzed With SwissADME.

Compound	Molecular weight(g/mol)	Log P	nHBA	TPSA	MR	Lipinski Violation	Log S	nRotB
Ovalichromene B	350.24	3.54	5	53.99	95.69	2	−4.58	1
Gamatin	336.29	3.20	6	71.04	90.25	0	−4.48	2
Pongaglabrone	306.27	3.23	5	61.81	83.75	0	−4.43	1
Pongapin	336.29	3.25	6	71.04	90.25	0	−4.61	2

Abbreviations: Log P: Log of octanol/water partition coefficient; Log S, log of solubility; MW, molecular weight; nHBA, number of hydrogen bond acceptor(s); nHBD, number of hydrogen bond donor(s); MR, molar refractivity; nRotB, number of rotatable bonds; TPSA, total polar surface area.

The in silico predictions of absorption, distribution, metabolism, and excretion (ADME) are presented in Table 4. The results indicated that all compounds are predicted to show high gastrointestinal absorption and be able to permeate the blood–brain barrier, an important property for compounds acting on the central nervous system to treat Alzheimer's disease. Moreover, various cytochromes (CYPs) vital for drug biotransformation, including CYP1A2, CYP2C19, CYP2C9, CYP2D6, and CYP3A4, are known to affect drug metabolism.²⁶ Ovalichromene B inhibited all cytochromes and P-glycoprotein (P-gp) substrate, while gamatin, pongaglabrone, and pongapin inhibited most cytochromes, but not P-gp substrate. These results show that all 4 compounds are promising potential drugs targeting AChE to treat Alzheimer's patients.

Table 4.

Acetylcholinesterase (ADME) Predictions of Isolated Compounds Computed by SwissADME.

Compound	Log K_p(cm/s)	GIAbs	BBBPer	Inhibitor Interaction
Compound	Log K_p(cm/s)	GIAbs	BBBPer	P-gpSubstrate	CYP1A2	CYP2C19	CYP2C9	CYP2D6	CYP3A4
Ovalichromene B	−5.85	High	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Gamatin	−5.58	High	Yes	No	Yes	No	Yes	Yes	Yes
Pongaglabrone	−5.75	High	Yes	No	Yes	Yes	Yes	Yes	Yes
Pongapin	−5.81	High	Yes	No	Yes	No	Yes	Yes	Yes

Abbreviations: BBB Per: blood–brain barrier permeability; CYP, cytochrome-P; GI Abs: gastrointestinal absorption; log Kp, log of skin permeability; P-gp, P-glycoprotein.

Conclusion

We constructed a deep learning model based on a dataset of public compounds. The model showed 94.3% accuracy, 97.1% precision, 95.9% recall, and 86.1% specificity. This model could be applicable for identifying chemical compounds that act as AChE inhibitors. The accuracy of our model was evaluated with a list of natural compounds extracted from P. pinnata. Four flavonoid compounds (1-4) selected from the deep learning model showed an inhibitory effect with experimental IC₅₀ values between 19.8 and 63.5 μg/mL. The interaction between these bioactive compounds and AChE was assessed with molecular docking. Most compounds had a noticeable interaction with the AChE catalytic residues Ser203 and His447. In ADME, 4 molecules possessed acceptable drug-like properties and druglikeness parameters. These results show that our deep learning model is suitable for future AChE inhibitor development studies, and P. pinnata is a potentially promising medicinal plant for treating Alzheimer's disease.

Experimental

Deep Learning Model

Data Preparation

A drug's IC₅₀ of AChE is widely used to measure its potency. In this study, the dataset was collected from the ChEMBL database and previous studies. Anonymous and null data were dropped. The chemical formula was saved in simplified molecular-input line-entry system (SMILES) format. Molecular descriptors of a drug included 881 features (molecular fingerprints). 881 Molecular descriptors were calculated from the SMILES formula by PaDeL-Discriptor. IC₅₀ values of AChE were categorized into “active” and “inactive” with a cut-off value of 50 μg/mL. Finally, the resultant data amounted to 5230 input vectors. Then, the dataset was divided into 80% and 20% as the training set and validation set, respectively (Figure 10).

Figure 10.

Data description and classes.

The list of compounds extracted from P. pinnata was collected from our previous studies^27,28 and these molecular features were also calculated by PaDeL-Discriptor. This dataset was used for the testing model.

Model Building

For prediction, input data consisted of chemical properties of compounds, and we split the dataset into training and validation sets. The CNN method was applied for prediction. Conv1D layers and max-pooling layers were used to import and reduce features of the dataset, then the input data were added to the neural network. The parameters are formed and improved through the training process in the training set and evaluated by the validation set. The model to predict AChE inhibitors was built by python language on Google Colaboratory.

AChE Inhibition Assay

The AChE activity was assayed by adapting a reported spectrophotometric method reported.²⁹ The assay was made in triplicate in 96-well microplates, and an ELISA microplate reader (Biotek, USA) was used to measure the absorbance. To assess the inhibitory impact of the samples on AChE activity, acetylthiocholine iodide (ACTI) was used as the substrate. The reaction mixture, containing 140 μL phosphate buffer (pH 8.0), 20 μL of AChE solution (0.25 IU/mL), and 20 μL of tested sample was incubated for 15 min at 25 °C. After that, 10 μL of 2.5 mM ACTI, and 10 μL of 2.5 mM Ellman's reagent (DTNB) was added to initiate the reaction. The mixture was incubated further at 25 °C for 15 min, and the absorbance at 405 nm was measured. Galanthamine was used as a positive control. All tested samples and the positive control were dissolved in 10% DMSO. The tested sample was measured at doses of 100, 20, 4, and 0.8 μg/mL and IC₅₀ values were calculated by the program TableCurve, Version 4.0.

Molecular Docking Simulation

Protein and Ligand Preparation

The crystal structure of AChE (PDB ID: 3LII) was retrieved from Protein Data Bank (RCSB PDB). Next, water molecules were removed and polar Kollman charges were added to the protein using AutoDock tools version 1.5.6. Finally, the macromolecule was exported into a dockable pdbqt format for molecular docking.

The information on compounds of P. pinnata was collected from previous studies and 3D structures were prepared by MarvinSketch (ChemAxon, USA)

The 3D structures of the compounds (1-4) isolated from P. pinnata were retrieved from either the PubChem database or prepared by MarvinSketch (ChemAxon, USA). After that, all ligands were converted to dockable pdbqt format utilizing AutoDock tools version 1.5.6.

Molecular Docking

The molecular docking process on AChE was carried out using AutoDock Vina. The grid box was identified with parameters: center_x = −11.2, center_y = −42.8, center_z = −27.3 and size_x = 22.50, size_y = 18.75, size_z = 20.25. Docking scores are reported in kcal/mol. Finally, Discovery Studio Visualizer was used to visualize the molecular interactions between proteins and ligands.

In Silico Druglikeness

Prediction of druglikeness of compounds was based on the already established concept of Lipinski's rule.³⁰ SwissADME server was used to calculate the ADME properties of ligands which provided information about the ADME properties of the ligands.³¹

Footnotes

Availability of Data and Materials

The dataset and the model used during the current study are available in the Github repository .

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Ethical Approval

Ethical Approval is not applicable for this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the University of Medicine and Pharmacy, Hue University (grant number 23/21).

ORCID iDs

Tan Khanh Nguyen

Thanh Hoa Tran

Duc Viet Ho

Linh Thuy Thi Tran

Statement of Human and Animal Rights

This article does not contain any studies with human or animal subjects.

Statement of Informed Consent

There are no human subjects in this article and informed consent is not applicable.

References

Khachaturian

. An overview of Alzheimer’s disease research. Am J Med. 1998;104(4 A):26S‐31S. doi:10.1016/s0002-9343(98)00026-6

McGleenon

Dynan

Passmore

. Acetylcholinesterase inhibitors in Alzheimer’s disease. Br J Clin Pharmacol. 1999;48(4):471‐480. doi:10.1046/j.1365-2125.1999.00026.x

Mehta

Adem

Sabbagh

. New acetylcholinesterase inhibitors for Alzheimer’s disease. Int J Alzheimer Dis. 2012;2012:1-8. doi:10.1155/2012/728983

Muñoz

Inestrosa

. Neurotoxicity of acetylcholinesterase amyloid β-peptide aggregates is dependent on the type of Aβ peptide and the AChE concentration present in the complexes. FEBS Lett. 1999;450(3):205‐209. doi:10.1016/S0014-5793(99)00468-8

Zhavoronkov

. Artificial intelligence for drug discovery, biomarker development, and generation of novel chemistry. Mol Pharm. 2018;15(10):4311‐4313. doi:10.1021/acs.molpharmaceut.8b009306

Khanh

Hoa

. In silico studies of natural products from medicinal plants to identify potential inhibitors for SARS-CoV-2 3C-like protease. Vietnam J Chem. 2021;59(5):557‐562. doi:10.1002/vjch.202000140

Tran

LTT

, et al. Virtual screening and in vitro evaluation to identify a potential xanthine oxidase inhibitor isolated from Vietnamese Uvaria cordata. Nat Prod Commun. 2022;17(2):1‐8. doi:10.1177/1934578X221080339

Tran

LTT

Dang

NYT

Nguyen

, et al. In silico and in vitro evaluation of alkaloids from Goniothalamus elegans Ast. for breast cancer treatment. Nat Prod Commun. 2022;17(3):1‐10. doi:10.1177/1934578X221088110

Sharma

Jindal

. Machine learning and deep learning applications-a vision. Global Transitions Proc. 2021;2(1):24‐28. doi:10.1016/j.gltp.2021.01.004

10.

Everingham

Eslami

SMA

Van Gool

Williams

CKI

Winn

Zisserman

. The pascal visual object classes challenge: A retrospective. Int J Comput Vision. 2015;111(1):98‐136. doi:10.1007/s11263-014-0733-5

11.

Russakovsky

Deng

, et al. Imagenet large scale visual recognition challenge. Int J Comput Vision. 2015;115(3):211‐252. doi:10.1007/s11263-015-0816-y

12.

Chan

Samala

Hadjiiski

Zhou

. Deep learning in medical image analysis. Adv Exp Med Biol. 2020;1213:3‐21. doi: 10.1007/978-3-030-33128-3_1

13.

Ghumare

Jirekar

Farooqui

Naikwade

. A review of Pongamia pinnata-an important medicinal plant. Curr Res Pharm Sci. 2014;04(02):44‐47.

14.

Punitha

Vasudevan

Manoharan

. Effect of Pongamia pinnata flowers on blood glucose and oxidative stress in alloxan induced diabetic rats. Indian J Pharmacol. 2006;38(1):62‐63. doi:10.4103/0253-7613.19858

15.

Shoba

Thomas

. Study of antidiarrhoeal activity of four medicinal plants in castor-oil induced diarrhoea. J Ethnopharmacol. 2001;76(1):73‐76. doi:10.1016/S0378-8741(00)00379-2

16.

Lale

Kulkarni

. A mosquito repellent Karanj Kunapa from Pongamia pinnata. Asian Agrihist. 2010;14(2):207‐211.

17.

John Samuel

AJS

Radhamani

Gopinath

Kalusalingam

Anbumani Vimala

AGK

Husain

. In vitro screening of anti-lice activity of Pongamia pinnata leaves. Korean J Parasitol. 2009;47(4):377‐380. doi:10.3347/kjp.2009.47.4.377

18.

Verma

Pradhan

Sharma

Naik

Prasad

. Efficacy of karanjin and phorbol ester fraction against termites (Odontotermes obesus). Int Biodeterior Biodegradation. 2011;65(6):877‐882. doi:10.1016/j.ibiod.2011.05.007

19.

Mamun

Shahjahan

Ahmad

. Laboratory evaluation of some indigenous plant extracts as toxicants against red flour beetle, Tribolium castaneum Herbst. J Bangladesh Agril Univ. 1970;7(1):1‐5. doi:10.3329/jbau.v7i1.4789

20.

Saini

Lakshmayya

Bisht

. Anti-Alzheimer activity of isolated karanjin from Pongamia pinnata (L.) Pierre and embelin from Embelia ribes Burm. f. Ayu. 2017;38(1):76. doi:10.4103/ayu.ayu_174_16

21.

Khan

Marya Amin

Kamal

Patel

. Flavonoids as acetylcholinesterase inhibitors: current therapeutic standing and future prospects. Biomed Pharmacother. 2018;101(October 2017):860‐870. doi:10.1016/j.biopha.2018.03.007

22.

Uriarte-Pueyo

Calvo

. Flavonoids as acetylcholinesterase inhibitors. Curr Med Chem. 2011;18(34):5289‐5302. doi:10.2174/092986711798184325

23.

Johnson

Moore

. The peripheral anionic site of acetylcholinesterase: structure, functions and potential role in rational drug design. Curr Pharm Des. 2006;12(2):217‐225. doi: 10.2174/138161206775193127

24.

Dvir

Silman

Harel

Rosenberry

Sussman

. Acetylcholinesterase: from 3D structure to function. Chem Biol Interact. 2010;187(1-3):10‐22. doi: 10.1016/j.cbi.2010.01.042

25.

Jogoth

Patrick

Marc

Andrew

Stewart

. Revisiting the general solubility equation: in silico prediction of aqueous solubility incorporating the effect of topographical polar surface area. J Chem Inf Model. 2012;52(2):420‐428. doi: 10.1021/ci200387c

26.

Ulrich

Matthias

. Cytochrome P450 enzymes in drug metabolism: regulation of gene expression, enzyme activities, and impact of genetic variation. Pharmacol Ther. 2013;138(1):103‐141. doi: 10.1016/j.pharmthera.2012.12.007

27.

Hoai

Hanh

HTN

Duc

Hoa

Tai

. Furanoflavonoids from the plant Pongamia pinnata (L.) (Fabaceae) growing in the province of Thua Thien Hue (Central Vietnam). Pharm J. 2014;54:42‐46.

28.

Duc

Hanh

HTN

Hoai

Hoa

. Pyranoflavanone and sterol from the plant Pongamia pinnata (L.) (Fabaceae) growing in Thua Thien Hue Province. J Med Mater. 2014;19(2):101‐106.

29.

Ellman

Courtney

Andres

Featherstone

. A new and rapid colorimetric determination of acetylcholinesterase activity. Biochem Pharmacol. 1961;7:88‐95. doi: 10.1016/0006-2952(61)90145-9

30.

Benet

Hosey

Ursu

Oprea

. BDDCS, the Rule of 5 and drugability. Adv Drug Deliv Rev. 2016;101:89‐98. doi: 10.1016/j.addr.2016.05.007

31.

Daina

Michielin

Zoete

. SwissADME: a free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Sci Rep. 2017;7:1‐13. doi: 10.1038/srep42717