Tài liệu Tương lai của sàng lọc ảo trong khám phá thuốc chữa bệnh

Nội dung text: Tài liệu Tương lai của sàng lọc ảo trong khám phá thuốc chữa bệnh

1. 42 Pham Thi Ly, Le Quoc Chon / Tạp chí Khoa học và Công nghệ Đại học Duy Tân 02(45) (2021) 42-50 02(45) (2021) 42-50 The future of virtual screening in drug discovery Tương lai của sàng lọc ảo trong khám phá thuốc chữa bệnh Pham Thi Lya, Le Quoc Chona,b* Phạm Thị Lya, Lê Quốc Chơna,b* aFaculty of Pharmacy, Duy Tan University, Da Nang, 550000, Vietnam. aKhoa Dược, Trường Đại học Duy Tân, Đà Nẵng, Việt Nam bInstitute of Research and Development, Duy Tan University, Da Nang, 550000, Vietnam bViện Nghiên cứu và Phát triển Công nghệ Cao, Trường Đại học Duy Tân, Đà Nẵng, Việt Nam (Ngày nhận bài: 26/01/2021, ngày phản biện xong: 06/02/2021, ngày chấp nhận đăng: 03/03/2021) Abstract One of the key elements in the early stages of drug discovery is finding good hits to develop lead compounds. Although HTS has been used as a standardized technology for hit finding, it still bears some challenging drawbacks: expensive and low-quality data. Aiming at the same goal as HTS, virtual screening (VS) has been developed to reduce cost and increase efficiency. Recent studies show that VS can deliver numerous quality hits and a few of them even reach clinical trials. This paper uses HTS as a background to discuss the contributions, limitations and research trends in VS field as these two technologies complement each other. Keywords: high-throughput screening, virtual screening, hit compounds, lead compounds, structure-based drug design, ligand-based drug design. Tóm tắt Bước quan trọng đầu tiên của quá trình khám phá thuốc chữa bệnh là tìm ra các hợp chất chất dẫn. Và phương pháp sàng lọc hiệu suất cao (HTS) được xem là công nghệ chuẩn để thực hiện bước này. Tuy nhiên, sử dụng HTS còn gặp nhiều khó khăn do chi phí cao và hiệu suất thấp. Với cùng mục đích sử dụng, sàng lọc ảo (VS) được phát triển để khắc phục các nhược điểm của HTS, để giảm chi phí và tăng hiệu quả tìm kiếm các chất dẫn. Những nghiên cứu mới đây cho thấy VS có thể cung cấp các chất dẫn có chất lượng cao, một số đã và đang được thử nghiệm lâm sàn. Bài báo này thảo luận các đóng góp, hạn chế và xu hướng phát triển của VS trong tương lai. Từ khóa: sàng lọc hiệu suất cao, sàng lọc ảo, hợp chất dẫn, khám phá thuốc, thiết kế thuốc dựa trên máy tính. 1. Introduction identifying drug targets and this task is Drug discovery and development (DDD) is routinely carried out by high-throughput a risky, and expensive process (figure 1). screening (HTS), filtering and selecting the Screening to find hits is conducted after most suitable molecules among a library of * Corresponding Author: Le Quoc Chon; Institute for Research and Training in Medicine, Biology and Pharmacy, Duy Tan University, Da Nang; Department of Pharmacy, Duy Tan University, Da Nang, 550000, Vietnam Email: lequocchon@duytan.edu.vn
2. Pham Thi Ly, Le Quoc Chon / Tạp chí Khoa học và Công nghệ Đại học Duy Tân 02(45) (2021) 42-50 43 small chemicals or biologics [1]. For over two technology in pharma and biotech companies decades, HTS has become a standard [2], [3]. Figure 1: Seven steps in drug discovery HTS technology has evolved over three on efficiency and the third has emphasized on generations [4]: the first focused on quantity of the flexibility and quality of library (figure 2). compound screened, the second concentrated Figure 2: Workflow of contemporary HTS [5]. HTS has been substantially contributing to number is unknown, the chemical space may DDD through providing new chemical entities contain about 1060 molecules [18]. While a good for lead development [6]–[9]. Leveraging library must be diverse and lead-like [16], a automation and miniaturization technologies, it typical library for HTS holds a few ten thousand has accelerated the drug discovery process [10] to less than two million compounds [2]. and generated many FDA-approved drugs [9]. The next disadvantage of HTS is high cost. Three critical factors determining the success of Running HTS is expensive and time-consuming HTS campaign are druggable target, compound because it is experimental-based. A single HTS library and predictive assay [11], [12]. screening program costs approximate 75 000 USD Even though HTS has been widely used, the [13]. HTS service cost ranges from 0.1 to 1 USD number of drugs approved by US-FDA is per well [19]. Furthermore, the robotic systems constantly low over two decades [13]. Given and assay readers in HTS are costly, requiring up that HTS is a routine operation in to a few millions of USD to set-up and maintain pharmaceutical [14], it has been partly to blame [2], [20]. In addition to resource, conducting HTS for the decline in DDD [9], [15]. campaign is time-consuming, taking several Indeed, HTS has some inherent drawbacks. months to a year to finish [21]. Furthermore, HTS First of all, the compound library is too small in program is highly specific, automatizing some comparison to the possible chemical space [16]. expensive steps as assay development and Medicinal chemistry assumes that there are validation is impossible. Moreover, HTS needs sufficient small molecules for all binding sites real high-quality library, containing many drug- found in biology [17]. Even though the exact like molecules to screen, and an enormous amount
3. 44 Pham Thi Ly, Le Quoc Chon / Tạp chí Khoa học và Công nghệ Đại học Duy Tân 02(45) (2021) 42-50 of time and money is demanded for collecting, problem, and “Pan-Assay Interference” synthesizing or/and buying. compounds [22], [23] and frequently reports In spite of being equipped with advanced false-positive [24] and false-negative [25], [26]. technologies, HTS is possibly already reaching Moreover, data handling and analysis are its limited capacity. Currently, ultra-HTS can challenging in addition to the artifacts from sample 100 000 compounds per day using 384 readout technologies [27]. or 1536-well microplates. In addition, this 2. The role of virtual screening (VS) capacity depends on the availability of real One approach to alleviate the ugliness of compounds for cell-based or biochemical HTS is to embrace virtual screening (VS), assays, which is not always the case. which complements HTS, reducing the number Finally, HTS often reports biased results. It of compounds to be tested experimentally. encounters frequent hitters, aggregation Figure 3. Overview of two categories in virtual screening [28]. Two major VS categories (figure 3) have are known based on data from X-ray diffraction been used over two decades are ligand-based and NMR [29] or relied on homology modeling virtual screening (LBVS) and target-based [29]–[31]. LBVS, on the other hand, is used virtual screening (TBVS) [28], [29]. TBVS is when active ligands are known or the structure conducted when structures of target molecules of targets is established [32].
4. Pham Thi Ly, Le Quoc Chon / Tạp chí Khoa học và Công nghệ Đại học Duy Tân 02(45) (2021) 42-50 45 Many scientists have tried to figure out the A similar evaluation method is comparing contributions of VS [33]–[37], but it has not VS results with that of HTS [49]. Doman et al. been easy because it takes many years to see if for instance found that hit rate of VS was 1700- the hits found by VS turn up to be approved fold higher than HTS [38]. Similarly, Paiva et drugs. In addition, this field is still evolving and al. reported that hit rates from VS was 30-fold not all VS results are in public domain, higher than HTS [50]. Most recently, Damm- particularly those from pharmaceutical Ganamet et al. reported that hits found by HTS industry. To circumvent this difficulty, for overlapped with more than 70% found by VS instance, Van Vlijmen et al. had assessed the [40]. However, these groups used different contribution of computational chemist in drug libraries for screening, the comparison is discovery based on the number of parents that therefore imperfect. these chemists hold [37], while others have The second question concerns the quantity mentioned highly potent and diverse and quality of hits found. Despite the great chemotypes as successful proofs [33], [34], effort made, few hits discovered by VS have [36], [38]–[41]. become drug candidates and approved for To offer a global view on how VS patients [34]. So far, structure-based discovery contributes to DDD, Slater suggested that VS has helped to bring approximate 20 drugs in performance can be evaluated by focusing on clinical uses [51]. retrospective studies to extract statistical data [42]. Typically, about 0.1 – 2.5% top-ranked At smaller and specific case, VS molecules in VS result are selected as hit compounds, containing new chemotypes, on performance can be assessed based on the which potency can be improved for further ability to find hit compounds from a vast exploration [18], [52], [53]. From 2014 to 2018, chemical space. In the best scenario hit PubMed indexes about more than 100 compounds become approved drugs. Two publications using VS, altogether filtered over questions guide us to assess different aspects of 93 million compounds and found more than VS: (i) are VS methods reliable? (ii) is VS 12500 hits and bioassays confirmed that more efficient than other approaches in approximately 10% of these hits were actually searching for hits? bioactive [42]. The first question asks for the validation of The success of VS in drug discovery can be computational methods used in VS. The most demonstrated occasionally through some common way based on a reference library, triumphant stories. For example, compound containing known active compounds and PRX-08066 (1 in figure 4), a potent and decoys [43], [44]. If VS reproduces the selective antagonist at serotonin 5-HT2B experimental result, the method is accepted. In receptor against pulmonary arterial addition, the binding modes of the hits found hypertension [54], was discovered and designed must be experimentally verified [40], [45], [46] with the aid of computational approaches at and likewise, the biological activities must be EPIX Pharmaceuticals [34]. This compound reassessed through many bioassay formats in has 5-HT2B binding affinity (Ki) of 3.4 nM and laboratory setting to ensure the reproducibility has been also studied for cancer treatment [55]. of results [47], [48]. Similarly, Damm-ganamet et al. has found a
5. 46 Pham Thi Ly, Le Quoc Chon / Tạp chí Khoa học và Công nghệ Đại học Duy Tân 02(45) (2021) 42-50 highly potent hit series which became leads, against Hsp90α protein and found many highly among them one lead - a quinoline tertiary active hits which were confirmed by alcohol (2 in figure 4), has been developed to subsequent in vitro assays. The most potent hits be a New Chemical Entity [40]. Another are compound 3, 4 and 5 (in figure 4) with example comes from study of Al-Sha et al., experimental IC50 of 3, 5, 6 nM, respectively. they screened more than 240000 molecules Figure 4. Some most potent hits found by VS. Most recently, Lyu et al. screened 130 and these two conditions uncover the inherent 170 million compounds against AmpC and D4 limitations of VS. dopamine receptor, finding a phenolate 3. Limitation of VS inhibitor of AmpC and after being optimized, it attained binding affinity of 77 nM, placing it at Although VS is a promising technique to the top most potent non-covalent inhibitors find hits and new chemotypes for lead known [18]. Likewise, Gahlawat et al. screened compounds, it needs to overcome two libraries of natural products, FDA-approved limitations related to algorithm and database to drugs and known inhibitors; they found many achieve its full potential. potential lead compounds [56]. The largest drawback of VS is its suboptimal Even though we are unable to assess algorithms. Hits reported from VS frequently holistically VS’s performance because of contain false-positive compounds [34], [57], biased publish data [57] or lack of information [58]. Because VS screens huge chemical from pharmaceutical industry [42], these few libraries, the number of false-positive can be successful stories clearly demonstrated that VS large, the cost of synthesis and in vitro tests to has highly potential in DDD providing that it confirm results can be prohibitive. One reason accesses to quality libraries and equipped with is attributed to the current algorithms, which do reliable computational methods. Unfortunately, not take into account the flexibility of target
6. Pham Thi Ly, Le Quoc Chon / Tạp chí Khoa học và Công nghệ Đại học Duy Tân 02(45) (2021) 42-50 47 molecules, knowing as a critical factor 4. Outlook determining the accurate docking pose and VS has made grand contributions to DDD binding affinity prediction, thus the ranking regardless of its many limitations. Fortunately, result in VS [42], [59]. Sometimes the this field is still evolving, especially to cope algorithm is too specific; it works effectively with big data in Chemistry [65]. Building larger for some systems but not the others [60], target databases to accelerate drug discovery preventing it from screening across ligand- process is especially useful for VS [66]. target systems. Recently, sequencing of human genome and The other limitation links with the quantity, pathogen [2], [67], [68] and the advancement of quality and accessibility of ligand and target high-throughput crystallography and NMR [42] databases [61], [62]. Presently, chemical have offered more targets for VS. databases for DDD are fragmentary, some are Similarly, constructing huge and focused in public domain, others are industrial. VS ligand libraries for VS is significant [62], [69]– currently is unable synchronize all these [71]. Compounds are sourced from natural sources of databases to access potential products [72], [73], combinatorial chemistry, compounds more efficiently. Furthermore, the DNA-encode library [74], [75]. These ligand databases are still too tiny in comparison databases can be stored on cloud-based system to the possible chemical space [62]. This to ease access [76]. Aiming at target screening, limitation asks for more effort to build standard focused chemical databases have been built and libraries with easily access [63]. If size of using lead-like and drug-like properties as well and accessibility to databases are important, so as target relevant characteristics as filtration is quality. Since VS cannot find good hits if the criteria before screening. Today, many libraries stock does not hold that molecules as (table 2) have been made available for public experiences demonstrated [64]. usage [77]–[79]. Table 2: Some big and popular libraries for virtual screening Library Source and purpose Number of compounds URL bioactive compounds with drug-like properties from ChEMBL 15 million medicinal chemistry literature PubChem compounds from academic screening centers 109 million ChemSpider compound structures from multiple sources 101 million commercially-available compounds for virtual ZINC 980 million screening SureChEMBL compounds from chemical patents 17 million Developing better computational methods to Nowadays, the most obvious trend in drug account for the flexibility of drug target is also discovery is the application of AI, specially an obvious trend [42], [80]. To simplify the machine learning and deep learning have complexity of the system, most of the already show many promising potentials [81], algorithms in the past studies ignore the [82] dealing successfully with huge chemical flexibility of target molecules and this is one of databases. the reasons the screening result is not satisfactory. The improved methods are expected to deliver accurate VS results.
7. 48 Pham Thi Ly, Le Quoc Chon / Tạp chí Khoa học và Công nghệ Đại học Duy Tân 02(45) (2021) 42-50 5. Conclusion [10] R. Macarron, “Critical review of the role of HTS in drug discovery,” Drug Discov. Today, vol. 11, no. Virtual screening has made important 7–8, pp. 277–279, 2006. contributions to DDD, providing many hit and [11] F. Prinz, T. Schlange, and K. Asadullah, “Believe it lead compounds with diverse chemical scaffold or not: How much can we rely on published data on potential drug targets?,” Nat. Rev. Drug Discov., for further exploitation. Hitherto, to achieve its vol. 10, no. 9, pp. 712–713, 2011, [Online]. full potential, VS must overcome two Available: challenges. First, it requires more reliable [12] Editorial, “Stay on target,” Nat. Chem. Biol., vol. 9, no. 4, p. 193, 2013. computational methods to account for the [13] J. Bajorath, “Integration of virtual and high- flexibility of drug targets and better scoring throughput screening,” Nat. Rev. Drug Discov., vol. functions. Second, VS demands big and high- 1, no. 11, pp. 882–894, 2002. [14] E. J. Gordon, “Small-Molecule Screening: It takes a quality chemical libraries to be used as Village,” ACS Chem. Biol., vol. 2, no. 1, pp. 9–16, 2007. reference database to validate computational [15] R. Lahana, “How many leads from HTS?,” Drug methods, simultaneously as a pool to search for Discov. Today, vol. 4, no. 10, 1999. [16] C. A. Lipinski, “Lead- and drug-like compounds: good hits. The screening efficiency would be The rule-of-five revolution,” Drug Discov. Today higher if the chemical libraries are synergized Technol., vol. 1, no. 4, pp. 337–341, 2004. and focused, and can be accessed easily. The [17] J. L. Reymond and M. Awale, “Exploring chemical space for drug discovery using the chemical future of VS relies on its capability to handle universe database,” ACS Chem. Neurosci., vol. 3, reliably big chemical databases using AI. no. 9, pp. 649–657, 2012. References [18] J. Lyu et al., “Ultra-large library docking for discovering new chemotypes,” Nature, vol. 566, no. [1] J. Inglese et al., “High-throughput screening assays 7743, pp. 224–229, 2019, [Online]. for the identification of chemical probes,” Nat. [19] UWCCC Research, “Services, Equipment & Chem. Biol., vol. 3, no. 8, pp. 466–479, 2007. Pricing,” University of Wisconsin-Madison. [2] L. M. Mayr and P. Fuerst, “The future of high- throughput screening,” J. Biomol. Screen., vol. 13, equipment-services/ (Jan. 2021). no. 6, pp. 443–448, 2008. [20] M. Zaragoza-Sundqvist, H. Eriksson, M. Rohman, [3] J. Noah, “New developments and emerging trends and P. J. Greasley, “High-quality cost-effective in high-throughput screening methods for lead compound management support for HTS,” J. compound identification,” Int. J. High Throughput Biomol. Screen., vol. 14, 2009. Screen., p. 141, 2010. [21] U. of Helsinki, “High through­put chem­ical [4] D. A. Pereira and J. A. Williams, “Origin and evolution of high throughput screening,” Br. J. screen­ing.” Pharmacol., vol. 152, no. 1, pp. 53–61, 2007. [5] L. M. Mayr and D. Bojanic, “Novel trends in high- discovery-chemical-biology-and- throughput screening,” Curr. Opin. Pharmacol., vol. screening/infrastructures/high-throughput- 9, no. 5, pp. 580–588, 2009. biomedicine/(accessed Jan. 21, 2021). [6] A. Golebiowski, S. R. Klopfenstein, and D. E. [22] N. Thorne, D. S. Auld, and J. Inglese, “Apparent Portlock, “Lead compounds discovered from activity in high-throughput screening: origins of libraries: Part 2,” Curr. Opin. Chem. Biol., vol. 7, compound-dependent assay interference,” Curr. no. 3, pp. 308–325, 2003. Opin. Chem. Biol., vol. 14, no. 3, pp. 315–324, 2010. [7] S. Fox, H. Wang, L. Sopchak, and R. Khoury, [23] J. L. Dahlin et al., “Post-HTS case report and “High-Throughput Screening: Early Successes structural alert: Promiscuous 4-aroyl-1,5- Indicate a Promising Future,” J. Biomol., vol. 6, no. disubstituted-3-hydroxy-2H-pyrrol-2-one actives 3, pp. 137–140, 2001. verified by ALARM NMR,” Bioorganic Med. Chem. [8] A. Davis, D. Keeling, J. Steele, N. Tomkinson, and Lett., vol. 25, no. 21, pp. 4740–4752, 2015, [Online]. A. Tinker, “Components of Successful Lead [24] A. Jadhav et al., “Quantitative analyses of Generation,” Curr. Top. Med. Chem., vol. 5, no. 4, aggregation, autofluorescence, and reactivity pp. 421–439, 2005. artifacts in a screen for inhibitors of a thiol [9] R. MacArron et al., “Impact of high-throughput protease,” J. Med. Chem., vol. 53, 2010. screening in biomedical research,” Nat. Rev. Drug [25] S. C. Owen, A. K. Doak, P. Wassam, M. S. Discov., vol. 10, no. 3, pp. 188–195, 2011, [Online]. Shoichet, and B. K. Shoichet, “Colloidal
8. Pham Thi Ly, Le Quoc Chon / Tạp chí Khoa học và Công nghệ Đại học Duy Tân 02(45) (2021) 42-50 49 Aggregation A ff ects the E ffi cacy of Anticancer [40] K. L. Damm-Ganamet et al., “Accelerating Lead Drugs in Cell Culture,” 2012. Identification by High Throughput Virtual [26] W. P. Walters and M. Namchuk, “Designing Screening: Prospective Case Studies from the screens: How to make your hits a hit,” Nat. Rev. Pharmaceutical Industry,” J. Chem. Inf. Model., vol. Drug Discov., vol. 2, no. 4, pp. 259–266, 2003. 59, no. 5, pp. 2046–2062, 2019. [27] I. Caraus, A. A. Alsuwailem, R. Nadon, and V. [41] J. A. Bikker and L. S. Narasimhan, “Lead Discovery Makarenkov, “Detecting and overcoming systematic Using Virtual Screening,” Top. Med. Chem., vol. 5, bias in highthroughput screening technologies: A pp. 85–124, 2009. comprehensive review of practical issues and [42] O. Slater and M. Kontoyianni, “The compromise of methodological solutions,” Brief. Bioinform., vol. virtual screening and its impact on drug discovery,” 16, no. 6, 2015. Expert Opin. Drug Discov., vol. 14, no. 7, pp. 619– [28] D. Stumpfe and J. Bajorath, “Current trends, 637, 2019, [Online]. overlooked issues, and unmet challenges in virtual [43] K. E. Hevener et al., “Validation of Molecular screening,” J. Chem. Inf. Model., vol. 60, no. 9, pp. Docking Programs for Virtual Screening against 4112–4115, 2020. Dihydropteroate Synthase,” J. Chem. Inf. Model. [29] Z. Chen et al., “Pharmacophore-based virtual Model., vol. 49, no. 2, 2010. screening versus docking-based virtual screening: A [44] L. Chaput, J. Martinez-Sanz, N. Saettel, and L. benchmark comparison against eight targets,” Acta Mouawad, “Benchmark of four popular virtual Pharmacol. Sin., vol. 30, no. 12, pp. 1694–1708, 2009. screening programs: construction of the [30] A. Evers and G. Klebe, “Successful virtual active/decoy dataset remains a major determinant of screening for a submicromolar antagonist of the measured performance,” J. Cheminform., vol. 8, no. neurokinin-1 receptor based on a ligand-supported 1, pp. 1–17, 2016. homology model,” J. Med. Chem., vol. 47, no. 22, [45] G. L. Warren et al., “A critical assessment of pp. 5381–5392, 2004. docking programs and scoring functions,” J. Med. [31] R. Rodriguez, G. Chinea, N. Lopez, T. Pons, and G. Chem., vol. 49, no. 20, pp. 5912–5931, 2006. Vriend, “Homology modeling, model and software [46] C. K. Chang et al., “Structure-based virtual evaluation: Three related resources,” screening and experimental validation of the Bioinformatics, vol. 14, no. 6, 1998. discovery of inhibitors targeted towards the human [32] J. Y. Lee, J. M. Krieger, H. Li, and I. Bahar, coronavirus nucleocapsid protein,” Mol. Biosyst., “Pharmmaker: Pharmacophore modeling and hit vol. 12, no. 1, pp. 59–66, 2016. identification based on druggability simulations,” [47] W. L. Ye et al., “Virtual screening and experimental Protein Sci., vol. 29, no. 1, 2020. validation of eEF2K inhibitors by combining [33] D. E. Clark, “What has computer-aided molecular homology modeling, QSAR and molecular docking design ever done for drug discovery?,” Expert Opin. from FDA approved drugs,” New J. Chem., vol. 43, Drug Discov., vol. 1, no. 2, pp. 103–110, 2006. no. 48, pp. 19097–19106, 2019. [34] D. E. Clark, “What has virtual screening ever done [48] D. Lowe, “How to Know When a New Target is for drug discovery?,” Expert Opin. Drug Discov., Really a New Target,” Science Translational vol. 3, no. 8, pp. 841–851, 2008. Medicine, 2017. [35] G. Schneider, “Virtual screening: an endless staircase?,” Nat. Rev. Drug Discov., vol. 9, pp. 273– 06/30/how-to-know-when-a-new-target-is-really-a- 276, 2010. new-target (accessed Dec. 11, 2020). [36] S. Kar and K. Roy, “How far can virtual screening [49] K. Babaoglu et al., “Comprehensive Mechanistic take us in drug discovery?,” Expert Opin. Drug Analysis of Hits from High-Throughput and Discov., vol. 8, no. 3, pp. 245–261, 2013. Docking Screens against β-Lactamase,” J. Med. [37] H. van Vlijmen, R. L. Desjarlais, and T. Chem., vol. 51, no. 8, 2008. Mirzadegan, “Computational chemistry at Janssen,” [50] A. M. Paiva, D. E. Vanderwall, J. S. Blanchard, and J. Comput. Aided. Mol. Des., vol. 31, no. 3, pp. J. W. Kozarich, “Inhibitors of dihydrodipicolinate 267–273, 2017. reductase, a key enzyme of the diaminopimelate [38] T. N. Doman et al., “Molecular docking and high- pathway of Mycobacterium tuberculosis,” Biochim. throughput screening for novel inhibitors of protein Biophys. Acta - Protein Struct. Mol. Enzymol., vol. tyrosine phosphatase-1B,” J. Med. Chem., vol. 45, 1545, no. 1–2, pp. 67–77, 2001. no. 11, 2002. [51] J. J. Irwin and B. K. Shoichet, “Docking Screens for [39] M. Seifert and M. Lang, “Essential Factors for Novel Ligands Conferring New Biology,” J. Med. Successful Virtual Screening,” Mini-Reviews Med. Chem., vol. 59, no. 9, pp. 4103–4120, 2016. Chem., vol. 8, no. 1, pp. 63–72, 2008. [52] R. D. Cramer, “Leadhopping – and beyond,” Expert Opin. Drug Discov., vol. 1, 2006.
9. 50 Pham Thi Ly, Le Quoc Chon / Tạp chí Khoa học và Công nghệ Đại học Duy Tân 02(45) (2021) 42-50 [53] V. Katritch et al., “Structure-based discovery of [68] H. Fang, L. Chen, and J. C. Knight, “From genome- novel chemotypes for adenosine A2A receptor wide association studies to rational drug target antagonists,” J. Med. Chem., vol. 53, no. 4, pp. prioritisation in inflammatory arthritis,” Lancet 1799–1809, 2010. Rheumatol., vol. 2, no. 1, pp. e50–e62, 2020, [54] S. L. Porvasnik et al., “PRX-08066, a novel 5- [Online]. Available: hydroxytryptamine receptor 2B antagonist, reduces monocrotaline-induced pulmonary arterial [69] L. Ruddigkeit, R. Van Deursen, L. C. Blum, and J. hypertension and right ventricular hypertrophy in Reymond, “Enumeration of 166 Billion Organic rats,” J. Pharmacol. Exp. Ther., vol. 334, no. 2, pp. Small Molecules in the Chemical Universe Database 364–372, 2010. GDB-17,” J. Chem. Inf. Model., vol. 52, pp. 2864– [55] B. Svejda et al., “The 5-HT2B receptor plays a key 2875, 2012. regulatory role in both neuroendocrine tumor cell proliferation and the modulation of the fibroblast [70] D. E. Gloriam, “Bigger is better in virtual drug component of the neoplastic microenvironment,” screens,” Nature, vol. 566, 2019. Cancer, vol. 116, no. 12, pp. 2902–2912, 2010. [71] D. E. Clark, “Virtual screening: Is bigger always [56] A. Gahlawat et al., “Structure-Based Virtual better? Or can small be beautiful?,” J. Chem. Inf. Screening to Discover Potential Lead Molecules for Model., vol. 60, no. 9, pp. 4120–4123, 2020. the SARS-CoV-2 Main Protease,” J. Chem. Inf. [72] J. Shen et al., “Virtual Screening on Natural Model., 2020. Products for Discovering Active Compounds and [57] D. Stumpfe, P. Ripphausen, and J. Bajorath, “Virtual Target Information,” Curr. Med. Chem., vol. 10, no. compound screening in drug discovery,” Future Med. 21, pp. 2327–2342, 2005. Chem., vol. 4, no. 5, pp. 593–602, 2012. [73] J. M. Rollinger, H. Stuppner, and T. Langer, [58] Y. O. Adeshina, E. J. Deeds, and J. Karanicolas, “Virtual screening for the discovery of bioactive “Machine learning classification can reduce false natural products,” Prog. Drug Res., vol. 65, pp. positives in structure-based virtual screening,” Proc. 212–249, 2008. Natl. Acad. Sci. U. S. A., vol. 117, no. 31, pp. [74] R. A. Goodnow, C. E. Dumelin, and A. D. Keefe, 18477–18488, 2020. “DNA-encoded chemistry: Enabling the deeper [59] S. Saikia and M. Bordoloi, “Molecular Docking: sampling of chemical space,” Nat. Rev. Drug Challenges, Advances and its Use in Drug Discovery Discov., vol. 16, no. 2, 2017. Perspective,” Curr. Drug Targets, no. 5, 2019. [75] A. L. Satz, “What Do You Get from DNA-Encoded [60] E. Lionta, G. Spyrou, D. Vassilatis, and Z. Cournia, Libraries?,” ACS Med. Chem. Lett., vol. 9, no. 5, pp. “Structure-Based Virtual Screening for Drug 408–410, 2018. Discovery: Principles, Applications and Recent [76] C. Grebner, E. Malmerberg, A. Shewmaker, J. Advances,” Curr. Top. Med. Chem., vol. 14, no. 16, Batista, A. Nicholls, and J. Sadowski, “Virtual pp. 1923–1938, 2014. screening in the cloud: How big is big enough?,” J. [61] X. Zhang, S. Betzi, Z. Morelli, and P. Roche, “Focused chemical libraries – design and Chem. Inf. Model., vol. 60, no. 9, pp. 4274–4282, enrichment: an example of protein–protein 2020. interaction chemical space,” Future Med. Chem., [77] C. Gorgulla et al., “An open-source drug discovery vol. 6, no. 11, pp. 1291–1307, 2014. platform enables ultra-large virtual screens,” [62] W. P. Walters, “Virtual Chemical Libraries,” J. Nature, vol. 580, no. 7805, pp. 663–668, 2020. Med. Chem., vol. 62, no. 3, 2019. [78] A. Mullard and A. Mullard, “Large Pharmas Set Up [63] F. I. Saldívar-González, C. S. Huerta-García, and J. L. Screening,” vol. 18, 2019. Medina-Franco, “Chemoinformatics-based enumeration [79] J. J. Irwin and B. K. Shoichet, “ZINC – A Free of chemical libraries: a tutorial,” J. Cheminform., vol. Database of Commercially Available Compounds 12, no. 1, pp. 1–25, 2020, [Online]. for Virtual Screening,” J. Chem. Inf. Model., vol. [64] Editorial, “Screening we can believe in,” Nat. 45, no. 1, 2005. Chem. Biol., vol. 5, no. 3, p. 2009, 2009. [80] F. Spyrakis and C. N. Cavasotto, “Open challenges [65] I. V. Tetko, O. Engkvist, U. Koch, J. L. Reymond, in structure-based virtual screening: Receptor and H. Chen, “BIGCHEM: Challenges and modeling, target flexibility consideration and active Opportunities for Big Data Analysis in Chemistry,” site water molecules description,” Arch. Biochem. Mol. Inform., vol. 35, 2016. Biophys., vol. 583, pp. 105–119, 2015, [Online]. [66] Z. R. Tanoli et al., “Drug Target Commons 2.0: A [81] X. Yang, Y. Wang, R. Byrne, G. Schneider, and S. community platform for systematic analysis of drug- Yang, “Concepts of Artificial Intelligence for target interaction profiles,” Database, vol. 2018, no. Computer-Assisted Drug Discovery,” Chem. Rev., 2018, 2018. vol. 119, 2019. [67] L. Folkersen et al., “Genomic and drug target [82] N. Fleming, “How artificial intelligence is changing evaluation of 90 cardiovascular proteins in 30,931 drug discovery,” Nature, vol. 557, no. 7707, pp. individuals,” Nat. Metab., vol. 2, no. 10, pp. 1135– S55–S57, 2018. 1148, 2020, [Online].