Study on the Pharmacological Mechanism of Phthalazinone Derivatives as Potential Drugs for Alzheimer’s Disease Treatment by Network Pharmacology Analysis and Molecular Docking

Objective: To screen the bioactivity of phthalazinone derivatives for AD treatment and investigate the potential pharmacological mechanism, the network pharmacology analysis and molecular docking were adopted in this study. Methods: Those phthalazinone derivatives with certain structures and physical properties were screened out by Pubchem database in this study. Besides, to explore the potential activity of these phthalazinone derivatives as drugs for AD treatment, network pharmacology study was employed, including targets prediction, gene enrichment analysis and network analysis. Network analysis of AD approved drugs and molecular docking studies were also adopted to further investigate drug-likeness of phthalazinone derivatives for AD treatment. Results: Five compounds and 57 common targets were recognized and adopted to the construction of compounds-targets network. 15 approved drugs with clear indication for AD were gured out, with 57 associated targets that originated from Homo sapiens. The KEGG enrichment analysis showed that phthalazinone derivatives and approved drugs shared the same essential pathway (neuroactive ligand-receptor interaction) and other important pathways with associated targets. The result of molecular docking indicated that these phthalazinone derivatives could interact with essential targets stably. Conclusion: In silico analysis suggested that these derivatives are probable to be effective for the treatment of AD by interacting with the essential targets and initiating ATP binding, signal transduction, nally regulating neuroactive ligand-receptor interaction pathway (cid:0) calcium signaling pathway , and so on.

are, they could not be developed into drugs. Those compounds that conformed to the Lipinski rules and could be retrieved from Pubchem database would be employed for this study.

Prediction of Compound-Related Targets
The compound-related targets were predicted depending on chemical similarities and pharmacophore models visa Swiss Target Prediction (http://www.swisstargetprediction.ch/) database [28] , Batman TCM (http://bionet.ncpsb.org/batman-tcm/) database [29] , and PharmMapper (http://www.lilabecust.cn/pharmmapper/) database [30][31][32] . In the Batman database, those candidate targets with scores given by the target prediction method no less than 20 and p no more than 0.05 were recruited. Targets with Norm Fit more than 0.8000 in PharmMapper database, targets with probability rank top 10 and values greater than zero in Swiss Target Prediction database were also recruited. With the combination of all these targets, about 130 targets were available for later analysis.

Identi cation of AD-Related Targets
The known AD-related targets were extracted from TTD (Therapeutic Target Database, https://db.idrblab.org/ttd/) database [33] , and CTD (Comparative Toxicogenomics Database, http://ctdbase.org/) database [34] . Only those targets with inference score no less than 20 in CTD database and clinical trial targets for AD treatment in the TTD database were included. 2630 targets were found to conform to the condition without replicated targets.
2.4 Scanning of the common targets related to compounds and AD treatment All the targets obtained above were standardized as gene symbols and Uniprot IDs by searching from Uniprot (https://www.uniprot.org/) database with "Homo sapiens" species.
2.5 Scanning drugs approved for AD treatment Drugbank 5.1.7 (https://www.drugbank.ca/) and TTD database are usually used for the query of drugs and targets. The approved drugs with a clear indication for AD would be adopted, as well as their corresponding targets.

Network Construction and Topological Analysis
The network construction and topological analysis were mainly conducted by STRING 11.0 (https://string-db.org/cgi/input.pl) database and Cytoscape v3.7.2 software. The topological analysis was performed by the Cytoscape online network analyzer module. The parameters of Degree centrality (DC), betweenness centrality (BC), and closeness centrality (CC) would be displayed to evaluate the central properties of the nodes in the network. Those nodes with DC ≥ median DC, BC ≥ median BC, CC ≥ median CC, would be employed as the key targets of compounds.

GO and KEGG Pathway Enrichment analysis
The GO and KEGG enrichment analysis are of great importance for the study of functional annotation and pathways of selected genes. The GO and KEGG enrichment analysis were carried out by DAVID v6.8 (The Database for Annotation, Visualization and Integrated Discovery, https://david.ncifcrf.gov/) database [36,37] , with p-value set to less than 0.05.

Study of molecular docking
The study of molecular docking was performed by iGEMDOCK v2.1. The docking analysis of iGEMDOCK software was based on k-means and hierarchical clustering methods. The shared targets that were related to the same essential pathway or regarded as key targets of compounds would be extracted to molecular docking. As general cases, the default parameters for screening were adopted (population size = 200, generations = 70, number of solutions = 2).

Compounds for Analysis
In the Pubchem database, six compounds could be retrieved and ve of them accord with the Lipinski rules. The compounds' structures and other properties are listed in Table 1. All of the ve molecular have molecular weight more than 500, with other physical properties comply with Lipinski rules.

Compounds-common targets network of phthalazinone derivatives
The network of compounds-common targets was constructed by STRING database with minimum required interaction score of 0.400. The network ( Fig. 1) was consisted of 62 nodes (57 common targets and 5 compounds)with a centralization of 0.565. In this network, it is suggested that all the compounds recruited are possible to be effective for the treatment of AD, and compound 2 and compound 3 might enjoy more potential to be drugs for the treatment of AD, with a degree of 37, 43 respectively. The interactions between phthalazinone derivatives and common targets were as below. And all the compounds employed could interact with either PDE 4B or PDE4D.

PPI network of common targets
To further investigate the interaction between the compound targets, the PPI network was constructed. 57 targets were identi ed as common targets that both related to AD and phthalazinone derivatives, and 25 targets were regarded to directly interact with common targets. As shown in Fig. 2A, PDE 4D, NOS1, TNF, INS and other targets in pink were identi ed as common targets with higher degree value, which to some extent, mean that these targets enjoy more potential to be targets of phthalazinone derivatives to treat AD. According to the topological analysis of the network, INS, TNF, HSP90AA1, NOS1, PTGS2, ADRB2, HSPA1A, CALCA, ARRB1, TRPV1, NFKB1 and F2 were believed to be the key targets among common targets.
The common targets and associated targets were combined and employed to GO and KEGG enrichment analysis. The bubble diagram (Fig. 2B) suggested that the cGMP-PKG signaling pathway (hsa04022), Neuroactive ligand-receptor interaction (hsa04080) and Calcium signaling pathway (hsa04020) were more reliable to be considered as the major pathways that phthalazinone derivatives work on AD treatment. Besides the cAMP signaling pathway (hsa04024) as well as Morphine addition pathway (hsa05032) were also regarded as important pathway for phthalazinone derivatives. 3.4 Drug approved-targets network 15 kinds of drugs were gured out and approved with a clear indication for AD. The targets concerning were identi ed at the same time and employed to construct the drug approved-targets network. The network consisted of 72 nodes (57 targets and 15 drugs). It is apparent that these targets were prone to enjoy higher interaction (Fig. 3A). According to the topological analysis of this network, GRIN2B, GRIN1, GABRA1, GRIN2A, GABRB2, GLRB, GLRA2 and many other targets had no less than 30 neighbored nodes to interact with. The bubble diagram indicated that Neuroactive ligand-receptor interaction (hsa04080) and Nicotine addiction (hsa05033) pathways are of great essential for the approved drugs to take effect. Besides, Retrograde endocannabinoid signaling (hsa04723), GABAergic synapse (hsa04727) and Morphine addition (hsa05032) pathways were also considered as important in AD treatment (Fig. 3B).

Targets-pathways network
As what has been shown above, the KEGG pathways of phthalazinone derivatives and drugs approved for AD treatment were similar, sharing the same essential pathway, Neuroactive ligand-receptor interaction (hsa04080). To further compare the similarity of targets and pathways between phthalazinone derivatives and approved drugs, the targets-pathways networks were constructed. The pathways employed to analysis were conformed to the condition that p < 0.05 and associated gene count no less than 3.
Comparing the two gures below, it is notable that the network of approved drugs was more extensively interacted and each pathway enjoyed more gene count. However, the two networks also shared the same pathways and targets. For example, Neuroactive ligand-receptor interaction (hsa04080), Morphine addiction (hsa05032), cAMP signaling pathway (hsa04024), Calcium signaling pathway (hsa04020), Serotonergic synapse (hsa04726), and NF-kappa B signaling pathway (hsa04064) were the same in two networks. Besides, they shared the same targets, including GABRA2, CHRNA7, PTGS1, HTR1A, GABRB3, ADRA1A, NOS1, ADRA2A, PTGS2, SLC6A4 and NFKB1. Among them, the pathway of neuroactive ligandreceptor interaction was considered as essential in both the two networks. PTGS2, NOS1 and NFKB1 targets were also considered as key targets in the PPI network of phthalazinone derivatives; while GABRA2, CHRNA7, HTR1A, GABRB3, ADRA2A targets were found to be associated with neuroactive ligand-receptor interaction pathway. These eight targets would be considered as essential targets in this study.

Molecular docking study
The molecular docking result was calculated on the generic evolutionary method (GA) and presented by binding energy. Those with lower binding energy enjoy higher stability. The result indicated that phthalazinone derivatives could all interacted with these essential targets and keep more stable than associated bounded ligands (Fig. 5). NOS1 (4UCH) could be interacted and keep stable by compound 1, compound 3 and compound 4. In contrast with other phthalazinone derivatives, compound 1 and compound 4 enjoy higher binding capacity with the essential targets.

Phthalazinone derivatives' potential for AD treatment
AD, triggered by multi-targets, is prevalent among the aged and is of great necessity to develop drugs for treatment. Several traditional herbs and associated prescriptions have been employed to explore their potential mechanisms to treat AD by network pharmacology analysis and molecular docking study [38,39] . In this study, the in silico analysis all suggested that the phthalazinone derivatives employed were potential to be lead compounds to develop drugs for AD treatment. This nding conformed to the conclusions in published literatures. According to the available literatures, compounds with phthalazinone or phthalazine original nucleus were effective to treat AD, no matter on neuron cell groups or animal models [40] . For example, phthalazine derivates such as Hydralazine and 5IA, was found to be available to reduce oxidative damage and Aβ misfolding or function as an inverse agonist of 5 subunitcontaining Gamma-aminobutyric acid receptors (α5GABAARs) to retard the aggression of Aβ [41][42][43] .
What's more, Zopolestat, a phthalazinone derivative, had been reported to be available to inhibit Aβinduced neuroin ammation by regulating NF-κB and other signaling pathways, enjoying great potency to be drugs for AD treatment [44] .

Potential mechanism of phthalazinone derivatives
The gene enrichment analysis revealed that the candidate phthalazinone derivatives could interact with PDE4 and other targets, initiating cAMP signaling pathway, neuroactive ligand-receptor interaction, morphine addiction, and calcium signaling pathway (Fig. 4A, Fig. 2B), to take into effect in AD treatment. Such potential pathways not only were considered as reliable to explain the mechanisms of approved AD drugs (Fig. 3B, p < 0.01), but also could be extracted from the published literatures.
Neuroactive ligand-receptor interaction pathway could help to increase the release of dopamine, resulting in the protection of brain nerve [45] . Recent studies have reported that the opioid system was relevant to the development of various neurodegenerative diseases. The opioid drugs including morphine could induce intestinal dysbiosis and trigger changes in the central nervous system, interrupting the development of AD [46] . As what has been known, PDE 4 subtype is the cAMP-speci c phosphodiesterase and plays an essential role in the regulation of intracellular levels of cAMP. It is of great bene t to nourish the brain neuron and inhibit the aggression of Aβ by interrupting cAMP. By triggering cAMP/Protein Kinase A (PKA)/ cAMP-Response Element-Binding Protein (CREB)/ Brain-Derived Neurotrophic Factor (BDNF) signaling pathways, the drug could improve AD model mice's memory [47,48] . Shun Shimohama and Jun Kawamata indicated that the in ux of Ca 2+ would contribute to the intracellular signaling transduction and inhibit the aggression of Aβ [49] . That also means,calcium signaling pathway would in uence the development of AD. A literature review had already summarized that calcium channel could relate to all AD pathologies and calcium signaling pathway is a promising breakthrough for AD treatment [50] .
The GO and KEGG enrichment analysis (Fig. 2) suggested that the phthalazinone derivatives mainly took into effect by improving energy imbalance in the brains of AD patients.

Molecular docking analysis
There were two patterns for the preparation of protein (based on the current le and based on the bounded site) in molecular docking. The binding energy in these two patterns was nearly the same (STable 1). There was no bounded ligand in the target NFKB1 (2O61), and the pattern that preparing protein based on the current le was adopted to keep the consistency in this study. The binding energy among these phthalazinone derivatives were of no signi cant difference (Fig. 5). It is possible that the phthalazinone core structure with 2'--pyridine, and 8'--methoxyphenyl ensured these phthalazinone derivatives to interact with essential targets stably. The substituent group on pyridine made a little difference in the interaction. For example, hexatomic ring substituent enjoys better interaction stability than the furan-one. If there is an electrophilic group on the substituent and could interact with the receptor to form a hydrogen bond, this phthalazinone derivative would be easier to interact with.

Signi cance and application
Network pharmacology analysis was rst developed by Hopink in 2007, and today is popular in the study of the potential pharmacological mechanism of traditional Chinese medicine [51] . According to the methodology and theory of network pharmacology analysis, it could be adopted to initially screen the bioactivity of compounds and investigate the potential mechanism [22,52] . Different from other literatures about phthalazinones, this study employed network pharmacology analysis and molecular docking to suggest that those compounds could be effective in AD treatment. And network pharmacology analysis of the approved drugs was employed to further demonstrate the reliability of such bioactivity and mechanism. By network pharmacology analysis and molecular docking study, not only the potential targets and pathways, but also the binding capacity and structures-activity relationships were gured out.
The analytical pattern could be available to screen speci c bioactivity of the component group of traditional Chinese medicine and gure out the underlying pharmacological mechanism. It is also of great bene t to employing this pattern to discover new indications for the approved drug.
Network pharmacology analysis is a methodology based on the statistic posed on databases or published literatures, it is not a perfect match with the actual situation. And the targets we found on the database were displayed without actual interaction (agonist or antagonist) [53] . Pharmacodynamics research on AD animal models still needs to be conducted to verify the bioactivity of phthalazinone derivatives.

Conclusion
In this study, the analysis of phthalazinone derivatives suggested that these candidates were prone to enjoy bioactivity to treat AD and keep each interaction stably. The network pharmacology analysis of the approved drugs further testi ed the reliability of phthalazinone derivatives as drugs for AD treatment. All the analysis indicated that these phthalazinone derivatives were potential to be lead compounds to develop drugs for AD treatment. The GO enrichment analysis showed that the common targets could function involve in various biological processes, cellular components and molecular functions. The GO and KEGG enrichment analysis of common targets and associated targets suggested that these phthalazinone derivatives were effective mainly by interacting with signal transduction process, ATP binding, neuroactive ligand-receptor interaction, morphine addiction, cAMP signaling pathway, and calcium signaling pathway. The interactions above are conformed to the published literatures and could be employed to the treatment for AD. Network pharmacology analysis is usually applied to the potential pharmacological mechanism of traditional Chinese medicine. We think it is a good attempt to connect network pharmacology analysis and molecular docking with bioactivity screening. It is of great bene t to broaden the scale of network pharmacology applications.

Declarations
Ethics approval and consent to participate Not applicable. This manuscript does not report on or involve the use of any animal or human data or tissue.

Consent for publication
Not applicable. This manuscript does not contain data from any individual person.

Availability of data and materials
All data are available in the manuscript and are shown in tables, gures, and supplemental les.

Con ict of interest
There is no con ict of interest in this work.

Funding
Not applicable. There was no funding supported for this research.