A COMPUTATIONAL STUDY ON THE EFFICACY OF SMALL MOLECULES AS DUAL INHIBITORS FOR β-SECRETASE 1 AND ACETYLCHOLINESTERASE AS ALZHEIMER’S DISEASE THERAPEUTICS Текст научной статьи по специальности «Биологические науки»



ISSN 2310-5674 ppublishing.org

Section 3. Pharmaceutical science

DOI:10.29013/EJBLS-23-4-16-30

A COMPUTATIONAL STUDY ON THE EFFICACY OF

SMALL MOLECULES AS DUAL INHIBITORS FOR P-SECRETASE 1 AND ACETYLCHOLINESTERASE AS ALZHEIMER'S DISEASE THERAPEUTICS

Patrick Ming 1, Moustafa T. Gabr1

1 Princeton Day School, Weill Cornell Medicine, Cornell University

Cite: Patrick Ming, Moustafa T. Gabr. (2023). A Computational Study on the Efficacy of Small Molecules as Dual Inhibitors for fi-Secretase 1 and Acetylcholinesterase as Alzheimer's Disease Therapeutics. The European Journal of Biomedical and Life Sciences 2023, No 4 https://doi.org/10.29Q13/EJBLS-23-4-16-3Q

Abstract

Alzheimer's disease is characterized by the progressive decline of many cognitive functions that involve numerous parts of the brain. Due to the complex nature and multi-faceted pathogenesis of neurodegenerative diseases, AD's pathology has been correlated with a loss of cholinergic function as well as the overproduction of Afi42. Because of this, multitargeted ligands have great potential as therapeutics. In this paper, we assessed the "bindability" of the proteins using various computational methods and online software. Then, we virtually screened through millions of potential small molecules using pharmacophore maps to find potential small molecule candidates. We then used software developed by the Swiss Institute of Bioinformatics (SIB) to both assess the binding energy of the small molecules to the compound (of which all compounds had a AG < -7.00 kcal/ mol) as well as assess the druggability of the small compounds through their ADME profiles. By the end, we were left with 4 organic compounds that showed the most promise as dual inhibitors. Ranked from most to least promising, they are: ZINC68569271, ZINC41367268, ZINC67202317, ZINC05611209. They all show strong binding affinities for both AChE and BACE1, with the majority of each compound having a Gibbs free energy value of AG < - 7.00 kcal/mol. Keywords: Alzheimer's Disease, Acetylcholinesterase, fi-secretase 1, Molecular Docking, ADME

1. Introduction 1.1. Alzheimer's disease

Alzheimer's disease (AD) is characterized by the progressive onset of cognitive, functional, and behavioral impairment. Current-

ly, AD is the sixth leading cause of death and is one of the only top ten causes of deaths that is continually increasing in prevalence (Atri, 2019). AD is fatal due to a multitude of reasons, however one of the biggest prob-

lems with AD is identifying its existence. Unlike other diseases, AD's symptoms can often be attributed to regular aging. This causes symptoms to be left undiagnosed, untreated, and ignored, losing valuable time, and causing harmful delays in receiving appropriate therapeutics. For example, while lapses in memory and judgment are frequent symptoms among aging, complete irreversible loss of memory is a common side effect of AD (Atri, 2019). This demonstrates how AD often accentuates the symptoms of normal aging, often making it more severe.

As it stands right now it is estimated that by 2030, the number of Americans with AD would increase by 35%, with the potential to skyrocket up to tripling in prevalence by 2050. Current projections and estimates show that risk of dementia doubles roughly every 5 years from the age of 65 to 85. At 65 years old, the risk of developing AD is roughly 1% - 2%; by 85, the risk of developing neurodegenerative disease is 30% - 50% (Atri, 2019; Alzheimer's association, 2019).

1.2. AD Risk factors

Unlike other diseases, AD is thought to be multi-faceted, meaning its pathogenesis requires a combination of factors. So far, the greatest known risk factor for AD is age, followed closely by family history and one's genotype (Alzheimer's disease, 2019). Previous twin and family studies indicate that genetic factors play a major role in roughly 80% of AD cases (Tanzi, R. E., 2012; Gatz, M. et al., 2006).

O

Maintaining a healthy equilibrium of neurotransmitter systems such as ACh is essential to cognitive function. Although the

There are two categories of genes that can influence the pathogenesis of a disease: risk genes and deterministic genes. A risk gene simply increases the risk of developing a disease, while a deterministic gene all but guarantees the development of a disease. Rare mutations in the APP, PSEN1 and PSEN2 genes have been shown to be deterministic genes for AD, whereas common gene polymorphisms for the APOE gene such as s4 and s2 have been shown to influence but not cause the pathogenesis of AD (Tanzi, R. E., 2012).

1.3. Acetylcholine/Acetylcholines-terase

Neurons are the fundamental component of communication within an organic system. They communicate through electrochemical signals that are driven by charged particles. Communication occurs at tiny gaps between the neuron receptors called synapses, where the presynaptic neuron sends a neurotransmitter to the postsynaptic neuron which binds to receptor proteins on its surface (Lovinger, D. M., 2008). One of the most important neurotransmitters is called acetylcholine (ACh) (Kondziella, D., 2016), which is associated with AD and is one of the two proteins of focus of this study. Though AD has no cure, therapeutics can help delay symptoms or even completely halt the onset of dementia, one of the many symptoms of AD. Currently, the main class of drugs used to treat AD are acetylcholinesterase / cholin-esterase inhibitors (ChEIs).

cholinergic nervous system isn't the only neurological system affected by cognitive disorders such as AD, a lacking cholinergic

Figure 1. Catalyzed hydrolysis of ACh into HAc and Choline (Cavalcante, et al., 2020; Trang, A. et al., 2020)

system has been attributed to the progressive impairment in learning and memory (Maruc-ci, G. et al., 2020). For example, a premature loss of cholinergic neurons in the basal fore-brain was observed for those afflicted with AD. Acetylcholinesterase is an enzyme with the capability to catalyze the hydrolysis of acetylcholine into acetate and choline (see fig. 1), thus reducing its concentration at the synapse and consequently, its ability to help conduct a synapse (Trang, A. et al., 2020).

1.4. P-secretase

^-secretase (BACE1) is responsible for the aggregation of amyloid-P (AP) and is the second of the two proteins that we are studying (Ma, W.-H. et al., 2021). BACE1 is responsible for the aggregation of AP42, which has been correlated to increased toxicity in AD (Ham-pel, H., 2020). BACE1 has been measured in higher concentration and heightened activity in AD patients' brains and body fluids; corroborating the idea that BACE1 plays a critical role in the pathogenesis of AD (Hampel, H., 2020; Breijyeh, Z et al., 2020;).

Familial AD (FAD) cases are linked to mutations in the gene for amyloid precursor protein (APP) production. Patients who exhibit "increased dosage" of the APP gene, which is located on chromosome 21, are at higher risk to develop early-onset AD and overproduce the AP42 peptide (Vassar, R.

et al., 2009). Thus, the BACE1 enzyme is a prime drug target to develop inhibitors that inhibit the amyloid aggregation of AP42.

1.5. Tau and Amyloid Aggregation

Another field of AD therapeutic research

is the inhibition of tau and amyloid aggregation. Amyloid aggregation is the aggregation of the protein called amyloid in organs which can cause organ failure. As mentioned before, patients with FAD have a genetic mutation on their 21st chromosome which allows the overproduction of AP42 (Ma W.-H. et al., 2021; Hardy, J., 2006; Breijyeh, Z et al., 2020;).

Under normal physiological conditions, the tau (t) protein monitors and regulates the structural stability of microtubules and microfibers. However, in a diseased brain such as an Alzheimer's brain, tau becomes abnormally hyperphosphorylated. Then, the tau proteins aggregate together to form paired helical filaments, which then form neurofibrillary tangles which get in the way of the nervous system (Medeiros, R. et al.,)

1.6. FDA-approved therapeutics:

AChE has proven itself to be a viable target for therapeutic drugs and symptomatic improvement against the development of AD. Inhibition of AChE has real clinical results because of its proven efficacy as a treatment for previous ailments such as myasthe-nia gravis (Mehta M. et al., 2012).

Figure 2. Chemical diagrams of known AChEIs. Note physostig-mine is no longer in use (Marucci, G. et al., 2020)

Currently, AChEIs are the main class of drugs to treat AD, with other drugs making it onto the list (Marucci, G. et al., 2020). Currently, the FDA-approved drugs available to treat AD are galantamine, rivastigmine, do-nepezil, and memantine (Medications for Memory, Cognition and Dementia-Related Behaviors, 2021). Note that while the first three drugs are AChEIs, memantine is not, rather than inhibiting acetylcholinesterase, it inhibits glutamate, a neurotransmitter that brains which are afflicted with AD produce too much of; which can cause harmful side effects (NHS, 2021).

1.7. Dual inhibitors of BACE1 and AChE

Both the amyloidal aggregation caused by the BACE1 protein and lack of cholinergic function caused by AChE has made research into finding dual inhibitors for both enzymes a hot spot for scientists in the field. This process of targeting two or more bio-molecular targets with a single molecule is called the "multi-target directed ligands" (MTDL) process.

The MTDL approach is more well-received when attempting to therapize patients suffering from multi-faceted diseases such as AD. Reportedly, the MTDL approach is suggested to have seen better efficacy and safety compared to the original "STDL" approach (Stern, N. et al., 2022). Previous works by researchers discovered only 107 already FDA-approved drugs that can bind to AChE. Of those 107 drugs, 33 of them can also bind to BACE1 (Stern, N. et al., 2022).

2. Methods

2.1. Preparation of screening process and tools

Proteins and enzymes operate via binding sites; to properly inhibit their function, small molecules must preferentially bond to these binding sites to obstruct other objects from doing so. The approach we will use for this computational study is first, we identify readily available binding sites on both enzymes using freely available software online. Next, we aim to utilize pharmacophore maps to map out the interaction between our preferred enzyme and a receptor as the base pharmacophore map to compare other compounds too.

2.2. Identification of binding sites on AChE and BACE1

Computationally, we can identify binding sites on our enzymes of choice using readily available free software online. There are two processes by which we can operate: Firstly, we can preliminarily detect potential binding sites on the protein by using a geometric method to identify the sizes of these binding sites. Note that if the binding site is too large, there is no proper "cup" that can hold the small molecule in place. Conversely, if the binding site is too small, then the small molecule cannot even fit in the binding site. For our experiments, we will be looking at a range of 100-550 A, preferably looking for sites with volumes around 300 A. This will be done using online software developed by the University of Hamburg called Proteins Plus (Schoning-Stierand, K. et al., 2022; Schoning-Stierand, K. et al., 2020; Fahr-rolfes, R. et al., 2017).

The second process of experimentation we will use is an energy-based method. The potential energy of the interaction between our small molecule and the binding site is as critical to our assessment of the binding site as the geometry of it. If there is no favorable energetic interaction between the binding site and small molecule, then nothing will happen. We will assess the energy of our interactions using online software developed by the VAJDA lab at Boston University called FT Site. Our last process of experimentation is a machine learning-based method. Instead of looking at the geometry or the energy of the interactions, machine learning is used to predict the quality of the ligand binding site. We will utilize online software called PrankWeb, which builds upon P2Rank, a stand-alone command line program that predicts ligand binding sites from its structure (Prankweb, 2017; Jackubec, D. et al., 2022; Jendele, L. et al, 2019; Krivak, R., 2018; Cusbg, 2020;).

2.3. Virtual screening of small molecules

Though sources dispute on the exact value, it is commonly agreed that only under 10% of the human proteome is drug-gable (Kana, O., et al., 2019). This means that potentially either of our enzymes of interest are not druggable, i.e., they do not have any suitable binding sites. After having

identified whether or not BACE1 and AChE are druggable, we will move on to our next phase of experimentation, where we will virtually screen millions of compounds to identify those that show promise when binding to either protein.

We will be accomplishing this through utilizing pharmacophore maps, a pharmacophore is a map of the interaction between two compounds which aims to identify important features for receptor binding. When looking at an interaction, there are three necessary points of information that are need-to-know. Firstly, we need to know the quantity of interactions; we ask, "how many interactions are there?" Secondly, we look at the location of the interactions; we ask, "where do these compounds interact?" Lastly, we must also identify the type of interactions; we ask, "what kind of intermolecular forces are making this interaction happen?" A pharmacophore map answers all these questions graphically.

Computationally, we will be utilizing an online software called PocketQuery (Koes D. R. et al., 2011; Koes, D. R., 2023). Developed by the University of Pittsburgh, Pocket Query looks at the interaction between a protein and a small molecule and generates a pharmacophore map for it based on the conditions we set. These pharmaco-phore maps is the main tool that we will be using to virtually screen through millions of FDA-approved compounds. If the pharmaco-phore map of the original interaction overlaps with the pharmacophore map of another compound, then the molecule of interest can bind very similarly to the original overlap and thus is a potential inhibitor of it.

In this study, we will first construct the pharmacophore map of the interaction between BACE1 and a peptide inhibitor, as peptide inhibitors are already rather small; the peptide we will be looking at is OM99-2. The reason we are looking at BACE1 first is because, in comparison to AChE, BACE1 is much less readily druggable. Thus, it is the more challenging target; and, if compounds that can inhibit it are identified, there is a much higher chance that it can also bind AChE. Conversely, had we identified compounds that could inhibit AChE first, there would be a much lower chance that it would have the dual function to also inhibit BACE1.

2.4 Molecular docking

Up to this point, while we can determine how well compounds can match the phar-macophore of OM99-2 with BACE1, with our current methods we cannot determine how well these compounds can actually bind to BACE1 and AChE. This is where we will utilize molecular docking; molecular docking is a key tool in structural molecular biology and computer-assisted drug design (Morris G. M. et al., 2008).

Molecular docking is a process where the energy of an interaction between a ligand and a biomolecular target is estimated. In this computational study, we will be utilizing an online docking software called SwissDock (Grosdidier, A. et al, 2011a; Grosdidier, A. et al., 2011b). Since some of the compounds potentially are not within SwissDock's database, we generated mol2 files detailing the composition of the ligands and small molecules using the ZINC IDs we found from Pocket Query and ZINCPharmarcy.

2.5 Pharmacokinetics, and ADME

Pharmacokinetics is a subbranch of pharmacology where the interaction between an administered drug and the human body is studied. Generally, there are four main parameters to be examined in this field, which includes absorption, distribution, metabolism, and excretion (aptly named ADME) (Grogan, S. et al., 2023).

Absorption is the process by which a drug, either in tablet or capsule form, enters the circulation system of the human body. This parameter determines how quickly the drug is able to enter the bloodstream and reach its desired location. Absorption itself has a sub-parameter, called liberation; liberation is the process by which the drug is released from its pharmaceutical form (i.e. tablet/ capsule) (Grogan, S. et al., 2023).

Distribution is a parameter that describes how a substance can distribute itself throughout the human body. This parameter varies from substance to substance as substances have differing biochemical properties. Put simply, the process of distribution has two main factors: diffusion, and convection. These can be affected by the size, polarity, or binding ability of the substance (Grogan, S. et al., 2023).

Metabolism is the process by which the body processes and decomposes the drug. If

a drug is metabolized by the body too quickly before it reaches its designated compartmen-tal destination, it is ineffective and therefore has an inadequate ADME profile; however, if the drug is difficult to metabolize, it has the potential to have potently toxic effects from being in the body for too long. Most of the metabolism done by the body onto drugs is through the phase I (CYP450) and PHASE II (UGT) reactions in the liver. The phase I reaction oxidizes the molecules and converts them into polar metabolites that can then be bound to polymers in the phase II reaction through glucuronidation (Miyauchi, Y. et al., 2021). Most importantly, the combined efforts of these processes generally inactivate the medical capabilities of the drugs processed. This demonstrates that a balance between being metabolized quickly or slowly is an essential part of the ADME profile (Gro-gan, S. et al., 2023; Miyauchi, M. et al., 2021).

Excretion is the process by which the drug is flushed and eliminated from the body. Generally, this process is done by the kidneys, however, excretion can also occur via the lungs, skin, or the gastrointestinal tract. In the kidney, drugs can be excreted by the passive glomerular filtration or the active tubular secretion (Grogan, S. et al., 2023; Madison J. E. et al., 2008).

2.6 LipinskiS Rule of 5

There are many biochemical properties that predict whether a substance would have an adequate ADME profile for draggability. For example, the number of hydrogen bond donors and acceptors, molecular weight, calculated LogP (CLogP), etc. The number of hydrogen bond donors and acceptors dictates how easily the drug can bond to receptors. A higher amount of bond donors/acceptors means that the drug will attempt to bind to every available protein on its way to the designated compartmental destination and reduce its effectiveness. Additionally, the molecular weight of the drug must be relatively light to allow skin absorption, as larger molecules cannot pass the corneal layer (Bos J. D. et al., 2020).

For Lipinski's set of rules, aptly named Lipinskis rule of 5, it requires that the compound has no more than 5 hydrogen bond donors and no more than 10 hydrogen bond acceptors. It also requires that the molecular

weight of the molecule stays under 500 Dal-tons, which is equivalent to one atomic mass unit. And finally, a CLogP no more than 5 (Benet L. Z. et al., 2016). The calculated LogP is a measure of how hydrophobic or hydro-philic a compound is. If hydrophobic, then the compound is lipophilic, and vice versa; if hydrophilic, then the compound is lipophobic.

For our purposes, we will utilize an online software called SwissADME, utilizing the top five compounds that we determined from our data from SwissDock (Daina, A. et al., 2017). [41, We can reconstruct their chemical structure in the software and the software will run an algorithm to determine many factors, namely the hydrogen bond donors and acceptors, molecular weight, CLogP, as well as any other important factors, such as gastrointestinal (GI) absorption, and blood-brain-barrier (BBB) permeant.

3. Results

3.1 Identifying binding sites on AChE and BACE1 geometrically

Using the geometric based method, we are given both the volume of the binding site as well as the drug score, which indicates the quality of the binding site and how drugga-ble it is. On the tables below, the volumes and drug scores are detailed for the top five binding sites on both AChE and BACE1

Table 1. - Top 5 binding sites on AChE filtered for size

Pocket me (Ä3) g Score

P_2 399.60 0.81

P_3 355.84 0.74

P_4 352.42 0.70

P_5 285.17 0.63

P_6 284.94 0.61

From table 1, we can infer that AChE is a very promising drug candidate. With an average volume of 335.59 A3, the average binding site sits well within the range decided earlier. Furthermore, the average drug score of the 5 pockets is 0.698. In figure 3, we can see a cartoon visualization of AChE as well as the 5 binding sites identified.

Figure 3. Top 5 identified binding sites on AChE through the geometric method

Table 2. Top 5 binding sites on BACE1 filtered for size.

Pocket Volume (Ä3) Drug Score

P_4 305.39 0.73

P_5 291.68 0.66

P_6 225.62 0.66

P_7 292.61 0.51

P_8 171.87 0.50

Figure 4. Top 5 identified binding sites on BACE1 through the geometric method

From table 2, we can see that ftsite predicts that BACE1 is also druggable. There are also multiple binding sites all within the volume range earlier determined. However, note that the average binding site and drug score, with an average volume of 257.43 A3,

and an average drug score of 61.2, is generally lower than that of AChE. From these results, we can reasonably infer that AChE is a more viable binding target than BACE1 is.

3.3 Identifying binding sites on AChE and BACE1 based off machine learning.

While the previous method predicted the presence of binding sites solely based off the geometry and the 3D structure of the protein. PrankWeb predicts the presence of binding sites using a machine learning algorithm that has become well adapted to predicting the presence of binding sites. PrankWeb is an extension of the machine learning algorithm P2Rank, which is a ligand binding site

Figure 5. The black binding site pocket profiles for AChE (top) and BACE1 (bottom)

Pocket 1 A

PrsrUirr rank: 1

29 CS

PrutaabiliiY «fl.fi vtlt

AA caunb J*

C«nurvJi1MW OTtS

© ® (3) <s>)

Pocket 1 dk

>

P«l»l llttll il.+r

DKH

AA tsunl It

1 Til

fol f*Tj ® ®

Figure 5 was generated by the machine learning-based algorithm multiple binding sites were identified. The machine learning algorithm detected only 2 pockets on BACE1 while it detected 12 pockets for AChE, further supporting our claim that BACE1 is less readily bindable relative to AChE.

In figure 5, the colored amino acids represent the detected binding sites on the protein. With the black colored amino acids representing the binding site that is ranked the highest. By being ranked the highest, the algorithm predicts that these binding sites are the most promising binding sites in terms of draggability. Figure 6 displays the pocket profiles of both pockets; the pocket profile is a quick summary of important parameters of the pocket such as its rank, the pocket score, its probability score, etc. We can see that these pockets have scores

of 29.65 and 28.47 for AChE and BACE1 respectively, which is indicative of their capability as binding sites. Overall, this shows us

that while AChE is more readily bindable than BACE1 is, both protein targets do show promise of being drug targets.

Figure 6. Images of AChE (left) and BACE1 (right) as generated from the machine learning-based method

3.4 Pharmacophore maps

Now that AChE and BACE1 have both been assessed to be viable inhibitor targets, we must identify various compounds that have the potential to act as a ligand and inhibitor for these proteins. Since AChE has a "binding advantage," - being more capable of

being bound - over BACE1, we will examine BACE1 first. Utilizing the base pharmacophore map of the interaction between BACE1 and OM99-2, we selected the top 5 interactions between the protein and small molecule. Listed below in table 3 is summative data on the information given about the interactions.

Table 3. - Data on 5 different interactions between BACE1 and OM99-2

PDB Chain Size Distance Avg AG Residues Score

2ZHR C 1 0 -5.86 VAL 0.8654

2ZHR C 2 4.4547 -3.02 VAL, ASN 0.8118

2ZHR C 3 5.925 -3.15333 GLU, VAL, ASN 0.7919

2ZHR C 2 7.3035 -3.565 ALA, PHE 0.7849

2ZHR C 3 7.3035 -3.13 ALA, GLU, PHE 0.7627

Residue

Figure

VAL

Residue

Figure

VAL, ASN

GLU, VAL, ASN

ALA, PHE

ALA, GLU, PHE

Table 4. - The top 3 compounds that matched the pharmacophore map of its specific interaction the closest for all 5 interactions selected

AA Residue: VAL

1 st Choice

ZINC12661691 RMSD: 0.026

2 nd Choice

ZINC05611209 RMSD: 0.027

3 rd Choice

ZINC05611211 RMSD: 0.027

AA Residue: VAL ASN

ZINC43179028 RMSD: 0.036

AA Residue: GLU VAL ASN

ZINC57202317 RMSD: 0.045

AA Residue: ALA PHE

ZINC92602312 RMSD: 0.061

AA Residue: ALA GLU PHE

ZINC68569271 RMSD: 0.089

ZINC04403279 RMSD: 0.037

ZINC43167268 RMSD: 0.049

ZINC09360392 RMSD: 0.062

ZINC78555142 RMSD: 0.089

ZINC12534110 RMSD: 0.046

ZINC71589484 RMSD: 0.052

ZINC00293884 RMSD: 0.068

ZINC60372054 RMSD: 0.091

When selecting compounds to use for this study, PocketQuery provides us with a parameter called the root mean squared deviation (RMSD) value. The RMSD value is a quantitative measure of the similarity between superimposed atomic coordinates (Kufareva, I. et al., 2012). For our purposes, the RMSD value is used to determine how

closely the chosen compound's pharmaco-phore map matches the overlay of the original pharmacophore map (the one between BACE1 and OM99-2). To ensure accuracy that we choose compounds with the highest likelihood to properly bind BACE1 (and hopefully AChE), we reject any compounds with values of RMSD > 0.10.

Figure 7. Binding energies of identified small molecules with the enzymes AChE and BACE1

I Binding Energy far AChE

3.5 Molecular Docking

From the data gathered and summarized in figure 7, we can see a few clear candidates as our "best" compound. For example, we can see that 43167268 shows promise as a candidate with its AG for AChE of 10.3. For our other compounds, we can determine which are the top five candidates by looking at the largest minimum AG value for each pair of values. Using this method, we will rank the top 5 compounds by ranking them from the largest minimum value to the small-

■ Binding Energy for BACE1 est minimum value. This yields the ranking: 1. ZINC68569271, 2. ZINC41367268, 3. ZINC12661691, 4. ZINC67202317, 5. ZINC05611209.

3.6 ADME Profiles

In table 5, we can see the respective ADME profiles of all the 5 compounds chosen from the docking procedures. Of the 5 compounds examined, 4 successfully passed Lipinski's Rules, while ZINC12661691, with a molar mass (MM) = 518.52 g/mol, exceeded the MM limit of 500 g/mol.

Table 5. - ADME profiles of the top 5 identified compounds from SwissDock

H-bond H-bond Do- accep-nors tors Log P Molecular Weight GI Absorption BBB per-meant Lipinski's Rules?

68569271 3 3 3.69 375.49 g/mol High No Yes

41367268 1 7 2.03 337.31 g/mol High No Yes

12661691 2 4 4.92 518.52 g/mol High No No

67202317 2 2 3.75 379.56 g/mol High No Yes

05611209 1 5 3.08 409.25 g/mol High No Yes

Though SwissADME deems these compounds incapable of crossing the blood brain barrier (BBB), chemical modifications can be made to these compounds in the lab to allow them to cross the BBB. These findings act as a good starting point for the next series of trials and experiments needed to further validate the work of this paper.

4. Conclusion

Alzheimer's disease is a disease that afflicts millions throughout the world and has a heavy prevalence in our modern day. Alzheimer's is believed to be a multi-faceted disease, the facets we chose to attack in this study is by inhibiting the functions of the proteins AChE and BACE1. AChE's function in the brain is to catalyze the hydrolysis of ACh, a vital neurotransmitter, thereby reducing its concentration and therefore effectiveness at the synapse. BACE1's function is to cleave the APP gene, which causes the overproduction of A^42 peptides, which form long fibrils that interfere with neuronic communication and can lead the pathogenesis of AD.

We applied the computational methods listed above to search for dual inhibitors that could effectively inhibit the functions of both AChE and BACE1, which prevents the hydrolysis of ACh and the production of A^42 pep-tides, respectively. Because different binding targets have varying levels of druggability, we first assessed how capable AChE and BACE1 were of being bound through varying computational methods. Through these methods, we aimed to identify potential binding sites and utilize the data gathered to determine the general druggability of both proteins.

We first looked utilized a method called ProteinsPlus that uses a Difference of Gaussian (DoG) filter to detect potential binding pockets purely based on the physical structure of the protein. The method then predicts various parameters that describe the size, shape, and chemical properties of the pocket detected. We used this method to give a surface-level indication to the druggability of the proteins. Next, we utilized an online software called PrankWeb to further assess the drugga-bility of potential binding sites on AChE and BACE1. PrankWeb utilizes machine learning to predict the location and druggability of binding sites on proteins. From these meth-

ods, we were able to determine that BACE1 was much less readily bindable than AChE is.

Due to the fact that AChE has a "binding advantage" over BACE1, it seemed more practical to first identify compounds that could inhibit the function of BACE1 first, then see how well these identified compounds could inhibit AChE. To accomplish this, we utilized pharmacophore maps. Using the interaction between BACE1 and OM99-2 as a base, we identified compounds that overlayed well with the base pharmaco-phore map to virtually screen through millions of compounds. Then, we chose the top fourteen compounds that bound to BACE1 the most similarly to OM99-2. We then utilized docking software to predict the binding energies (measured in AG°) of the interaction of these fourteen compounds with both AChE and BACE1. From the data collected, we chose the top 5 most appealing compounds that showed the capability to bind both BACE1 and AChE. Finally, we assessed the ADME profiles of the top 5 chosen compounds using online methods to determine how effective these compounds could be as a drug. Of our top 5 compounds identified, all but ZINC12661691 had adequate ADME profiles. All 5 compounds bound to both protein targets with binding energies less than -7.00 kJ/mol, with ZINC43167268 binding to AChE with a binding affinity of AG° = -10.30 kJ/mol. Thus, of the millions of compounds initially screened, we are left with only 4 compounds (from most to least promising): ZINC68569271, ZINC41367268, ZINC67202317, ZINC05611209. Note, Swis-sADME assessed that all 4 small molecules were incapable of crossing the BBB, this is a crucial part of drug discovery for neuro-degenerative diseases, as being incapable of crossing the BBB effectively makes the drug useless; however, simple chemical modifications to the compounds can render them BBB-permeable.

Due to the nature of computational studies like this paper, there are several limitations that this paper has. For example, the computational methods used in this paper such as ftsite, PrankWeb, SwissDock, and SwissADME are all based upon prediction and have not been physically observed. The predicted binding affinities of these compounds potentially do

not reflect the real-world binding affinity. To further validate the work of this paper, there are numerous next steps that this line of work can take. Firstly, to alleviate the risk of using such an uncertain and small sample size, this computational experiment can be conducted several more times to identify new compounds that are able to bind to the protein targets as readily or more readily than the currently identified compounds. Once this is accomplished,

the binding affinities of these compounds should be experimentally determined within the lab. Then pre-clinical in vitro trials can be conducted to determine the efficacy these drugs are within an organic system; note that during these In vitro trials, chemical modifications are to be made that can render the small molecules BBB permeable. Finally, clinical in vivo trails can be conducted to determine the efficacy of these drugs in the human body.

References

Alzheimer's association. (2019). Causes and Risk Factors. Alzheimer's Disease and Dementia; Alzheimer's Association. URL: https://www.alz.org/alzheimers-dementia/what-is-alz-heimers/causes-and-risk-factors Atri, A. (2019). The Alzheimer's Disease Clinical Spectrum. Medical Clinics of North America,- 103(2).- P. 263-293. URL: https://doi.org/10.1016/j.mcna.2018.10.009 Benet, L. Z., Hosey, C. M., Ursu, O., & Oprea, T. I. (2016). BDDCS, the Rule of 5 and drugabili-ty. Advanced Drug Delivery Reviews,- 101.- P. 89-98. URL: https://doi.org/10.1016/j. addr.2016.05.007

Bos, J. D., & Meinardi, M. M. H. M. (2000). The 500 Dalton rule for the skin penetration of chemical compounds and drugs. Experimental Dermatology,- 9(3).- P. 165-169. URL: https://doi.org/10.1034/j.1600-0625.2000.009003165.x Breijyeh, Z., & Karaman, R. (2020). Comprehensive Review on Alzheimer's Disease: Causes and Treatment. Molecules,- 25(24).- P. 5789. URL: https://doi.org/10.3390/mole-cules25245789

Cavalcante, S. F. de A., Simas, A. B. C., Barcellos, M. C., de Oliveira, V. G. M., Sousa, R. B., Cabral, P. A. de M., Kuca, K., & Franca, T. C. C. (2020). Acetylcholinesterase: The "Hub" for Neurodegenerative Diseases and Chemical Weapons Convention. Biomolecules,- 10(3).-1BL. URL: https://doi.org/10.3390/biom10030414 Daina, A., Michielin, O., & Zoete, V. (2017). Swiss ADME: a Free web Tool to Evaluate phar-macokinetics, drug-likeness and Medicinal Chemistry Friendliness of Small Molecules. Scientific Reports,- 7(1). URL: https://doi.org/10.1038/srep42717 Fahrrolfes, R., Bietz, S., Flachsenberg, F., Meyder, A., Nittinger, E., Otto, T., Volkamer, A., & Rarey, M. (2017). ProteinsPlus: a web portal for structure analysis of macromolecules. Nucleic Acids Research,- 45(W1).- W337-W343. URL: https://doi.org/10.1093/nar/ gkx333

Gatz, M., Reynolds, C. A., Fratiglioni, L., Johansson, B., Mortimer, J. A., Berg, S., Fiske, A., & Pedersen, N. L. (2006). Role of Genes and Environments for Explaining Alzheimer Disease. Archives of General Psychiatry,- 63(2).- 168 p. URL: https://doi.org/10.1001/ archpsyc.63.2.168

Grogan, S., & Preuss, C. (2023, July 30). Pharmacokinetics. PubMed; StatPearls Publishing.

URL: https://www.ncbi.nlm.nih.gov/books/NBK557744 Grosdidier, A., Zoete, V., & Michielin, O. (2011a). Fast docking using the CHARMM force field with EADock DSS. Journal of Computational Chemistry,- 32(10).- P. 2149-2159. URL: https://doi.org/10.1002/jcc.21797 Grosdidier, A., Zoete, V., & Michielin, O. (2011b). SwissDock, a protein-small molecule docking web service based on EADock DSS. Nucleic Acids Research,- 39(suppl).- W270-W277. URL: https://doi.org/10.1093/nar/gkr366 Hampel, H., Vassar, R., Strooper, B. D., Hardy, J., Willem, M., Singh, N., Zhou, J., Yan, R., Van-mechelen, E., Vos, A. D., Nistico, R., Corbo, M., Imbimbo, B. P., Streffer, J., Voytyuk, I., Timmers, M., Monfared, A. A. T., Irizarry, M., Albala, B., & Koyama, A. (2020). The 3-Secre-

tase BACE1 in Alzheimer's Disease. Biological Psychiatry,- 0(0). URL: https://doi. org/10.1016/j.biopsych.2020.02.001 Hardy, J. (2006). Amyloid double trouble. Nature Genetics,- 38(1).- P. 11-12. URL:

https://doi.org/10.1038/ng0106-11 Jakubec, D., Skoda, P., Krivak, R., Novotny, M., & Hoksza, D. (2022). PrankWeb 3: accelerated ligand-binding site predictions for experimental and modelled protein structures. Nucleic Acids Research,- 50(W1).- W593-W597. URL: https://doi.org/10.1093/nar/gkac389 Jendele, L., Krivak, R., Skoda, P., Novotny, M., & Hoksza, D. (2019). PrankWeb: a web server for ligand binding site prediction and visualization. Nucleic Acids Research,- 47(W1).-W345-W349. URL: https://doi.org/10.1093/nar/gkz424 Kana, O., & Brylinski, M. (2019). Elucidating the druggability of the human proteome with eFindSite. Journal of Computer-Aided Molecular Design,- 33(5).- P. 509-519. URL: https://doi.org/10.1007/s10822-019-00197-w Koes, D. R. (2023). PocketQuery. Pocketquery.csb.pitt.edu. URL: http://pocketquery.csb.pitt. edu/pocket.html

Koes, D. R., & Camacho, C. J. (2011). Small-molecule inhibitor starting points learned from protein-protein interaction inhibitor structure. Bioinformatics,- 28(6).- P. 784-791. URL: https://doi.org/10.1093/bioinformatics/btr717 Kondziella, D. (2016). The Top 5 Neurotransmitters from a Clinical Neurologist's Perspective. Neurochemical Research,- 42(6).- P. 1767-1771.URL: https://doi.org/10.1007/ s11064-016-2101-z

Krivak, R., & Hoksza, D. (2018). P2Rank: machine learning based tool for rapid and accurate prediction of ligand binding sites from protein structure. Journal of Cheminformatics,-10(1). URL: https://doi.org/10.1186/s13321-018-0285-8 Kufareva, I., & Abagyan, R. (2012). Methods of protein structure comparison. Methods in Molecular Biology (Clifton, N.J.),- 857.- P. 231-257. URL: https://doi.org/10.1007/978-1-61779-588-6_10

Lovinger, D. M. (2008). Communication networks in the brain: neurons, receptors, neu-rotransmitters, and alcohol. Alcohol Research & Health: The Journal of the National Institute on Alcohol Abuse and Alcoholism,- 31(3).- P. 196-214. URL: https://pubmed. ncbi.nlm.nih.gov/23584863 Ma, W.-H., Chen, A.-F., Xie, X.-Y., & Huang, Y.-S. (2021). Sigma ligands as potent inhibitors of AP and ApOs in neurons and promising therapeutic agents of Alzheimer's disease. Neuropharmacology,- 190.- 108342 p. URL: https//doi.org/10.1016/j.neuro-pharm.2020.108342

Maddison, J. E., Page, S. W., & Dyke, T. M. (2008). Clinical pharmacokinetics. Small Animal Clinical Pharmacology,- P. 27-40. URL: https//doi.org/10.1016/b978-070202858-8.50004-x

Marucci, G., Buccioni, M., Ben, D. D., Lambertucci, C., Volpini, R., & Amenta, F. (2020). Efficacy of acetylcholinesterase inhibitors in Alzheimer's disease. Neuropharmacology,-190.- 108352 p. URL: https//doi.org/10.1016/j.neuropharm.2020.108352 Medeiros, R., Baglietto-Vargas, D., & LaFerla, F. M. (2010). The Role of Tau in Alzheimer's Disease and Related Disorders. CNSNeuroscience & Therapeutics,- 17(5).- P. 514-524. URL: https//doi.org/10.1111/j.1755-5949.2010.00177.x Medications for Memory, Cognition and Dementia-Related Behaviors. (2021). Alzheimer's Disease and Dementia. URL: https//www.alz.org/alzheimers-dementia/treatments/ medications-for-memory#:~: text=Donepezil%20(Aricept%C2%AE)%3A%20approved Mehta, M., Adem, A., & Sabbagh, M. (2012). New Acetylcholinesterase Inhibitors for Alzheimer's Disease. International Journal of Alzheimer's Disease, 2012.- P. 1-8. URL: https//doi.org/10.1155/2012/728983 Miyauchi, Y., Takechi, S., & Ishii, Y. (2021). Functional Interaction between Cytochrome P450 and UDP-Glucuronosyltransferase on the Endoplasmic Reticulum Membrane: One of Post-translational Factors Which Possibly Contributes to Their Inter-Individual

Differences. Biological and Pharmaceutical Bulletin,- 44(11).- P. 1635-1644. URL: https//doi.org/10.1248/bpb.b21-00286 Morris, G. M., & Lim-Wilby, M. (2008). Molecular docking. Methods in Molecular Biology (Clifton, N.J.),- 443.- P. 365-382. URL: https//doi.org/10.1007/978-1-59745-177-2_19 NHS. (2022, August 31). Common questions about memantine. Nhs.uk. URL: https//www. nhs.uk/medicines/memantine/common-questions-about-memantine/#:~:text=Me-mantine%20works%20by%20blocking%20the P2Rank framework. (2023, March 14). GitHub. URL: https//github.com/cusbg/ p2rank-framework

Schöning-Stierand, K., Diedrich, K., Ehrt, C., Flachsenberg, F., Graef, J., Sieg, J., Penner, P., Poppinga, M., Ungethüm, A., & Rarey, M. (2022). ProteinsPlus: a comprehensive collection of web-based molecular modeling tools. Nucleic Acids Research,- 50(W1).- W611-W615. URL: https//doi.org/10.1093/nar/gkac305 Stern, N., Gacs, A., Tátrai, E., Flachner, B., Hajdú, I., Dobi, K., Bágyi, I., Dormán, G., Lorincz, Z., Cseh, S., Kígyós, A., Tóvári, J., & Goldblum, A. (2022). Dual Inhibitors of AChE and BACE-1 for Reducing Aß in Alzheimer's Disease: From in Silico to In Vivo. International Journal of Molecular Sciences,- 23(21).- 13098 p. URL: https//doi.org/10.3390/ ijms232113098

Swiss Dock - The online docking web server of the Swiss Institute of Bioinformatics - Docking. (n.d.). Swissdock. ch. Retrieved October 29, 2023. From URL: http://swissdock.ch/ docking

Tanzi, R. E. (2012). The Genetics of Alzheimer Disease. Cold Spring Harbor Perspectives in Medicine,- 2(10).- a006296-a006296. URL: https//doi.org/10.1101/cshperspect. a006296

Trang, A., & Khandhar, P. B. (2020). Physiology, Acetylcholinesterase. PubMed; StatPearls Publishing. URL: https//www.ncbi.nlm.nih.gov/books/NBK539735/#:~:text=Acetyl-cholinesterase%20(AChE)%20is%20a%20cholinergic Vassar, R., Kovacs, D. M., Yan, R., & Wong, P. C. (2009). The-Secretase Enzyme BACE in Health and Alzheimer's Disease: Regulation, Cell Biology, Function, and Therapeutic Potential. Journal of Neuroscience,- 29(41).- P. 12787-12794. URL: https//doi.org/10.1523/ jneurosci.3657-09.2009 Zentrum für Bioinformatik: Universität Hamburg - Proteins Plus Server. (n.d.). Proteins. plus. https://proteins.plus

submitted 12.10.2023; accepted for publication 31.10.2023; published 26.01.2024 © Patrick Ming, Moustafa T. Gabr. Contact: patrickming88@gmail.com