Use of Artificial Intelligence in Drug Discovery and its Application in Drug Development
Shaikh Habeeba
Matoshri Institute of pharmacy, Dhanore Yeola Mahrashtra India.
*Corresponding Author E-mail: habibashaikh762@gmail.com
ABSTRACT:
AI makes use of systems and software that can read and learn from data in order to make independent judgments in order to achieve certain goals. Its applications in the pharmaceutical area are constantly being expanded, as discussed in this chapter. The McKinsey Global Institute predicts that significant developments in AI-guided automation would fundamentally alter society's work culture5,6.
Machine learning (ML) is an area of AI that employs statistical methods with the ability to learn with or without being explicitly programmed7,8. There are three types of machine learning: supervised, unsupervised, and reinforcement learning. Classification and regression methods are used in supervised learning, and the prediction model is built using data from input and output sources. Under the subgroup classification, supervised ML produces disease diagnosis, and under the subgroup regression, it produces drug efficacy and ADMET prediction9. Clustering and feature-finding methods are used in unsupervised learning to group and analyse data based purely on input data10. Disease subtype discovery via clustering and disease target discovery from feature-finding approaches can both be achieved using unsupervised ML11. Reinforcement learning is primarily driven by decision-making in a given environment and the execution of those decisions in order to maximise performance. De novo drug design under decision making and experimental designs under execution are two examples of outputs from this sort of ML, both of which may be achieved using modelling and quantum chemistry12. Deep learning (DL) is a subfield of machine learning that use artificial neural networks to adapt and learn from huge amounts of data13,14. The ability to discover new compounds that could potentially be new drugs, uncover or repurpose drugs that could be more potent when used individually or in combination, and improve the area of personalised medicine based on genetic markers could be provided by big data and associated data mining and algorithm methods. With the increasing amount of data and the continuous expansion of computer power, the emergence of DL was observed.
The big data and associated data mining and algorithm methods can provide us with the capacity to discover new compounds that can potentially be new drugs, uncover or repurpose drugs that could be more potent when used individually or in combination and improve the area of personalized medicine based on genetic markers14. The emergence of DL was observed with the increasing amount of data and the continuous growth of computer power.
Artificial intelligence in drug discovery:
The enormous chemical space, which contains over 1060 compounds, encourages the synthesis of several drug molecules15. The lack of new technology, on the other hand, affects drug development, making it a time-consuming and costly that can be addressed by using AI16. AI can determine hit and lead compounds, leading to faster drug target validation and structural design optimization17,18. Despite its benefits, AI has significant data challenges, including the data's scale, growth, diversity, and uncertainty. Pharmaceutical companies' drug development sets of data can contain millions of molecules, and typical machine learning algorithms may not be able to handle them. A computational model based on the quantitative structure-activity relationship (QSAR) can quickly predict huge numbers of compounds or simple physicochemical parameters like log P or log D. However, these models are a some way from predicting complicated biological features like chemical efficacy and side effects. Small training sets, experimental data error in training sets, and a lack of experimental validations all are problems which QSAR-based models face. To overcome these issues, recently developed AI technologies, such as deep learning (DL) and relevant modelling studies, can be used to evaluate the safety and efficacy of drug molecules using big data modelling and analysis. Merck sponsored a QSAR ML challenge in 2012 to investigate the benefits of DL in the pharmaceutical industries or in drug discovery process. DL models showed significant predictivity compared with traditional ML approaches for 15 absorption, distribution, metabolism, excretion, and toxicity (ADMET) data sets of drug candidates19,20.
By illustrating the distributions of molecules and their properties, the virtual chemical space provides a geographical map of molecules. The goal of the chemical space illustration is to gather positional information about molecules inside the space in order to search for bioactive compounds, and therefore virtual screening (VS) helps in the selection of relevant molecules for further testing. PubChem, ChemBank, DrugBank, and ChemDB are just a few of the chemical spaces that are accessible to the public.
Numerous in silico methods for virtual screening compounds from virtual chemical spaces, as well as structure and ligand-based approaches, provide better profile analysis, faster elimination of nonlead compounds, and drug molecule selection at lower cost21. To select a lead compound, drug design algorithms like coulomb matrices and molecular fingerprint recognition take in consideration of physical, chemical, and toxicological properties22.
To predict the desired chemical structure of a compound, a variety of parameters can be used, including predictive models, molecule similarity, the molecule generation process, and the use of in silico approaches23, 24. When 95 000 decoys were tested against these receptors, Pereira et al. presented a new system, DeepVS, for the docking of 40 receptors and 2950 ligands, which demonstrated remarkable performance. Another method used a multiobjective automated replacement system to examine the form similarity, biochemical activity, and physicochemical features of a cyclin-dependent kinase-2 inhibitor to enhance its potency profile25.
QSAR modelling tools, such as linear discriminant analysis (LDA), support vector machines (SVMs), random forest (RF), and decision trees, have been used to identify prospective drug candidates and have evolved into AI-based QSAR approaches that can be used to speed up QSAR analysis26,27. When King et al. examined the capacity of six AI systems to rank anonymous compounds in terms of biological activity to that of traditional techniques, they found a negligible statistical difference28.
Artificial intelligence in drug screening:
A drug discovery and development can take more than a decade and cost an average of US$2.8 billion. Even still, nine out of 10 medicinal compounds fail to make it through Phase II clinical trials and receive regulatory approval29,30. Algorithms like Nearest-Neighbour classifiers, RF, extreme learning machines, SVMs, and deep neural networks (DNNs) are used to predict in vivo activity and toxicity in VS based on synthesis feasibility31,32. Several biopharmaceutical companies, including Bayer, Roche, and Pfizer, have collaborated with IT companies to establish a platform for drug discovery in areas including immuno-oncology and cardiovascular diseases. The aspects of VS to which AI has been applied are discussed below.
Prediction of the physicochemical properties:
Physicochemical properties of drug, such as solubility, partition coefficient (logP), degree of ionisation, and intrinsic permeability, have an indirect effect on its pharmacokinetic properties and target receptor family, and must be considered while developing a new drug33. Physicochemical properties can be predicted using a variety of AI-based methods. For example, to train the programme, ML makes advantage of vast data sets produced from previous compound optimization34.Molecular descriptors, such as SMILES strings, potential energy measurements, electron density around the molecule, and atom coordinates in 3D, are used in drug design algorithms to generate feasible compounds using DNN and predict their attributes35.
Kumar et al. used 745 compounds to train six predictive models [SVMs, ANNs, k-nearest neighbouring algorithms, LDAs, probabilistic neural network algorithms, and partial least square (PLS)] that were then applied to 497 compounds to predict their intestinal absorptivity based on parameters such as molecular surface area, molecular mass, total hydrogen count, molecular refractivity, molecular volume, logP, the sum of E-state indices, the solubility index (log S), and rotatable bonds36. In a similar lines, RF and DNN-based in silico models for predicting human intestinal absorption of a variety of chemical substances have been developed37. As a result, AI plays an important role in drug development, predicting not only the drug's desired physicochemical properties, but also its desired bioactivity.
a) Prediction of bioactivity:
The affinity of drug molecules for the target protein or receptor determines their efficacy. Drug molecules that do not interact with or have a high affinity for the targeted protein will not be able to provide a therapeutic response. It's also possible that produced therapeutic molecules interact with unintended proteins or receptors, resulting in toxicity in rare cases. As a result, DTBA (drug target binding affinity) is critical for predicting drug–target interactions.AI-based methods can assess a drug's binding affinity by considering at the features or similarities between the drug and its target. To determine the feature vectors, feature-based interactions recognize the chemical moieties of the drug and the target. In a similarity-based interaction, on the other hand, the similarity between the drug and the target is considered, and it is assume for predicting drug target interactions. web applications such as ChemMapper and the similarity ensemble method (SEA) are available38. KronRLS, SimBoost, DeepDTA, and PADME are a few of the ML and DL algorithms that have been usedto determine DTBA. To determine DTBA, ML-based approaches such as Kronecker-regularized least squares (KronRLS) assess drug and protein molecular similarity. SimBoost, on the other hand, uses regression trees to predict DTBA and considering both feature-based and similarity-based interactions. SMILES drug features; ligand maximum common substructure (LMCS), extended connectivity fingerprint, or a combination of thereof can be considered39 that similar drugs will interact with similar targets38.
b) Prediction of toxicity:
To avoid hazardous effects, it is critical to predict the toxicity of any drug molecule. Preliminary studies using cell-based in vitro assays are frequently employed, followed by animal studies to determine a compound's toxicity, increasing the cost of drug research. LimTox, pkCSM, admetSAR, and Toxtree are some of the web-based solutions that can assist cut costs. Advanced AI-based techniques search for similarities among compounds or predict the compound's toxicity based on input features. The Tox21 Data Challenge, sponsored by the National Institutes of Health, the Environmental Protection Agency (EPA), and the United States Food and Drug Administration (FDA), evaluated several computational techniques for forecasting the toxicity of 12 707 environmental compounds and drugs41. An ML algorithm called DeepTox outperformed all methods by identifying static and dynamic features within the data. The safety target prediction of 656 marketed drugs was evaluated using SEA against 73 unintended targets that might cause adverse effects. eToxPred was developed using an ML-based approach and was used to assess the toxicity and synthesis feasibility of small organic molecules with a 72 percent accuracy42. Similarly, open-source toxicity prediction algorithms like TargeTox and PrOCTOR are used40.
c) Prediction of blood brain barrier permeability:
Using 67–199 descriptors, the penetration of the blood–brain barrier (BBB) can be predicted computationally with 75–97% accuracy (for penetrating and non-penetrating molecules) [158]. Although the accuracy for non penetrating molecules is lower (60–80 percent), Zhao et al. reported that this bias in statistical learning methods might be overcome by using recursive feature elimination to extract the features from only 19 molecular descriptors, including polarizability, polarity-related properties, hydrogen bond properties, volume, weight, surface area, bond rotations, and pKa.41. The molecules in their training set were classified with an accuracy of over 90%.
The accuracy of their model in predicting the BBB penetration of molecules in a test set as greater than 95%. A decision tree induction ML method was developed in another investigation to predict BBB permeability42. With a successful classification rate of 90%, this model has a greater accuracy.
AI in advancing pharmaceutical product development:
The subsequent incorporation of a novel drug molecule into a suitable dosage form with desirable delivery characteristics followed the discovery of a novel drug molecule. In this case, AI can take the role of the classic trial-and-error approach43. With the use of QSPR44, various computational tools can handle problems encountered in the formulation design area, such as stability concerns, dissolution, porosity, and so on. Decision-support tools employ rule-based systems to select the type, nature, and quantity of excipients based on the drug's physicochemical characteristics, but they use a feedback mechanism to constantly monitor and modify the process45.
Guo et al. combined Expert Systems (ES) and Artificial Neural Networks (ANN) to establish a hybrid system for the development of direct-filling hard gelatin capsules containing piroxicam which satisfy the dissolving profile specifications. Based on the input parameters, the MODEL EXPERT SYSTEM (MES) makes formulation development decisions and recommendations. To ensure hassle-free formulation development, ANN uses backpropagation learning to link formulation parameters to the desired response, which jointly regulated by the control module43.CFD can also be used to investigate how tablet geometry affects the dissolving profile46. The combination of these mathematical models with AI could be extremely useful in accelerating pharmaceutical product development.
AI in clinical trial design:
Clinical trials, which take 6–7 years and a significant financial investment, are developed to assess the safety and efficacy of a drug product in humans for a particular disease condition. However, just one out of every ten molecules that enter these trials gets cleared, resulting in a huge loss for the industry47. Inappropriate patient selection, a lack of technical requirements, and poor infrastructure all can lead to these failures. However, with the vast amount of digital medical data available, AI can be used to reduce these failures48.
One-third of the clinical trial's time has been spent recruiting patients. The recruitment of suitable patients can ensure the success of a clinical study, which otherwise results in 86 percent of failure cases49. AI can assist in selecting only a specific diseased population for recruitment in Phase II and III of clinical trials by using patient-specific genome exposure profile analysis, which can help in early prediction of the available drug targets in the patients selected50. Preclinical molecule discovery and prediction of lead compounds before the start of clinical trials using other aspects of AI, such as predictive ML and other reasoning techniques, assistance in the early prediction of lead molecules that would pass clinical trials whilst also taking into consideration the selected patient population. Dropout rates from clinical trials account for 30% of clinical trial failure, leading to additional recruiting requirements for the trial's completion, resulting in a waste of time and money. This can be avoided by constantly monitoring the patients and helping them in complying to the clinical trials protocol. AiCure developed mobile technology to monitor regular medication intake by schizophrenia patients in a Phase II trial, which increased patient adherence by 25% and ensured the clinical trial's success.
Applications of AI in drug development:
Finding successful new drugs is a demanding task, and it is the most difficult aspect of drug research. The huge extent of what is known as chemical space, which is estimated to be in the order of 1060 molecules51, is the source of this. AI-enabled technologies have evolved into versatile tools that can be used at various stages of drug development, including drug target identification and validation, drug design, drug repurposing, R&D efficiency, aggregating and analysing biomedicine information, and refining the decision-making process to recruit patients for clinical trials52,53. These AI applications have the potential to mitigate the inefficiencies and uncertainties that afflict traditional drug development methods while also reducing bias and human intervention54. Prediction of feasible synthetic routes for drug-like molecules55, pharmacological properties56, protein characteristics as well as efficacy, drug combination and drug–target interaction57, and drug repurposing58 are a few of the applications of AI in drug development. The discovery of novel biomarkers and therapeutic targets, personalised medicine based on omic markers, and discovering the connections between drugs and diseases all are examples of how omic analysis can be used to identify new pathways and targets59,60. DL has a track record of identifying promising drug candidates and properly predicting their qualities as well as potential toxicity hazards. Using AI methods61, it is now possible to avoid prior challenges in drug development, such as analysis of enormous datasets, laborious screening of compounds while reducing standard error, and requiring large amounts of R&D cost and time of over US$2.5 billion and more than a decade62. New research can be conducted using AI technology to aid in the identification of new drug targets, rational drug design, and drug repurposing63,64.
AI in understanding the pathway or finding molecular targets:
In the area of drug development, AI has revolutionized the methodologies for identifying disease pathways and targets. This was made achievable by combining genetic data, biochemical characteristics, and target tractability65. The plausibility of predicting therapeutic targets using a computational prediction application known as 'Open Targets' a platform consisting of gene disease association data was probed in one study, and it was discovered that animal models exhibiting a disease-relevant phenotype with a neural network classifier of >71 percent accuracy provided the most predictive power66. AI can be used in the drug development process. IBM Watson for Drug Discovery, an AI platform, has discovered five new RNA-binding proteins (RBPs) connected to the pathophysiology of amyotrophic lateral sclerosis, a neurodegenerative disease (ALS)67.
AI in finding the hit or lead:
Use of AI in the discovery of small drug-like molecules is focused with chemical space use. Because it is possible to computationally enumerate the potential organic compounds, chemical space provides the stage for identifying unique and high-quality molecules68. Furthermore, machine learning techniques and predictive model software assist in the identification of target-specific virtual molecules and their association with their particular targets, all while improving the safety and efficacy qualities. By decrease the quantity of synthesised compounds that are then examined in either in vitro or in vivo systems, AI systems can reduce attrition rates and R&D expenditure69. With the available data on small-molecule modulator probes or their structural biology, a variety of in silico techniques for profile selection, such as virtual ligand or structure-based design approaches, can be used. In cases where structural data is insufficient, DL comes in very handy. Phenotypic data, disease, biology, or molecule network-based algorithms can all be used in this way. Validated AI techniques can be used to improve drug development success rates, but AI techniques under development must be validated before being used in drug development process. The synthesis of selected compounds is the most important step in the medication development process. As a result, AI is useful because it can prioritise molecules based on their ease of synthesis or design tools that are effective for the best synthetic route68,70.
AI in synthesis of drug-like compounds:
Drug-like molecules are those that follow Lipinski's five-step rule: (l) 500< Da molecular weight; (ii) H-bond donors: < 5; (iii) H bond acceptor: < 10; and (iv) computed Log P (cLogP): <5. Retro synthesis is widely used by chemists for the synthesis of drug-like molecules. The target compounds are recursively analyzed and sequentially converted into smaller fragments or building blocks that can be easily acquired or made in the retrosynthetic approach. The next stage is to figure out how these fragments will be converted into target compounds. The second step is the most difficult because it is difficult for the human brain to sort through the vast number of relevant organic reactions in the literature to identify the most plausible ones. By filling the voids that cause high failure in predicted organic synthesis (often known as 'out of scope' molecules), AI may help predict the best sought-after reactions. Unpredictable steric and electronic effects, as well as an insufficient understanding of the reaction process, are the main causes of voids in organic synthesis. Several computer aided organic compound synthesis (CAOCS) systems are currently available to assist chemists choose the best synthesis method; however, CAOCS is not part of the computer-aided drug discovery (CADD) workflow71.
Predicting the mode-of-action of compounds using AI:
Those involved in the drug development process, particularly medicinal chemists, are excited by the idea of having an AI platform that can predict on- and off-target effects as well as in vivo safety profiles of compounds before they are created. Drug development time, R&D costs, and attrition rates are all reduced when such platforms are available. DeepTox (which predicts the toxicity of novel chemicals) and PrOCTOR (which predicts the probability of toxicity in clinical trials) are two examples of such platforms72. If a larger and more precise dataset on toxicity and therapeutic profile of a diverse set of compounds is made available, the predictive accuracy of these platforms canbe improved. However, this can only be accomplished if the industry is prepared to exchange data. SPiDER, a new AI tool designed as an alternative to chemoproteomics to improve natural products for drug discovery, was recently developed73. SPiDER was used to predict the molecular target of b-lapachone, a clinical-stage natural naphthoquinone with anticancer activity, as a proof-of-concept. B-lapachone was predicted to be an allosteric and reversible regulator of 5-lipoxygenase by the platform (5-LO). A 5-LO functional assay is used to verify the prediction. Another AI tool, read-across structure–activity relationships (RASAR)74, was shown to accurately predict the toxicity of unknown compounds by connecting molecular structures and toxic properties by mining a large database of chemicals.
AI in selection of a population for clinical trials:
An ideal AI tool for clinical trials may identify patients' disease, identify gene targets, and predict the effect of the drug designed, including on- and off-target effects. In a Phase II trial of schizophrenic patients, a novel AI platform called AiCure was developed as a mobile application to monitor medication adherence, and it was claimed that AiCure increased adherence with 25% compared to traditional modified directly observed therapy'75. The process of selection of patients for a clinical trial is critical. Examining the relationship between human-relevant biomarkers and in vitro phenotypes allows for a more predictable and measurable assessment of therapeutic response uncertainty in a specific patient. The advances in artificial intelligence (AI) technologies to detect and predict human-relevant disease biomarkers enables for the recruitment of a specific patient population in Phase II and III clinical trials. Use of AI predictive modelling in the selection of a patient population may improve clinical trial success rates76.
AI in drug repurposing:
The process of drug repurposing becomes more appealing and feasible using AI. Applying an existing treatment to a new disease has the advantage of allowing the new drug to proceed directly to Phase II studies for a different indication without having to go through Phase I clinical trials and toxicology testing again77. DL applications were used to develop in silico methods for predicting drug pharmacological properties and drug repurposing using transcriptomic data from various biological systems and conditions. Deep neural networks (DNNs) are a highly adaptive multilayer network consisting of connected and interacting artificial neurons that perform various data manipulations78. The methods described are based on high-level representations of data using DNNs. DNNs were seen in a study by Aliper et al. to be able to classify complex drug action mechanisms on the pathway level, enabling drugs to be categorized into therapeutic categories based on their functional class, efficacy, therapeutic use, and toxicity79.
AI in polypharmacology:
Because of a deeper understanding of pathological processes in diseases at the molecular level, the 'one-disease–multiple-targets' paradigm currently reigns over the 'one-disease–one-target' paradigm. Polypharmacology is the study of a single disease with several targets. Molecular pathways, crystal structures, binding affinities, drug targets, disease relevance, chemical properties, and biological activities can be found in databases like ZINC, PubChem, Ligand Expo, KEGG, ChEMBL, DrugBank, STITCH, BindingDB, Supertarget, and PDB, among others. To design polypharmacological agents, AI could be utilized to examine these databases. The authors developed a computational platform, DeepDDI, for better understanding of drug–drug interactions and associated mechanisms, as well as prediction of alternative drugs for intended clinical use without negative health effects82, which was recently published in the literature as a big success of an AI application in designing polypharmacological agents.
CONCLUSION:
The advancement of AI, along with its remarkable tools, is constantly aimed at reducing obstacles faced by pharmaceutical companies, affecting the drug development process as well as the total lifespan of the product, which may explain the rise in the number of start-ups in this sector. The contemporary healthcare sector is confronted with a number of complex challenges, such as increasing drug and treatment costs, and society need significant changes in this field. Personalized medications with the necessary dose, release characteristics, and other required aspects can be made according to individual patient demand using AI in pharmaceutical product manufacturing. Using the latest AI-based technologies will not only reduce the time it takes for products to reach the market, but it will also improve product quality and overall safety of the manufacturing process, as well as provide better resource utilization and cost-effectiveness, thereby increasing the importance of automation.
REFERENCE:
1. Ramesh A. Artificial intelligence in medicine. Ann. R. Coll. Surg. Engl. 2004; 86:334–338.
2. Miles J., Walker A. The potential application of artificial intelligence in transport. IEE Proc.-Intell. Transport Syst. 2006; 153:183–198
3. Yang Y., Siau K. MWAIS; 2018. A Qualitative Research on Marketing and Sales in the Artificial Intelligence Age.
4. Wirtz B.W. Artificial intelligence and the public sector—applications and challenges. Int. J. Public Adm. 2019; 42:596–615.
5. Smith R.G., Farquhar A. The road ahead for knowledge management: an AI perspective. AI Mag. 2000; 21 17–17.
6. Lamberti M.J. A study on the application and use of artificial intelligence to support drug development. Clin. Ther. 2019; 41:1414–1426.
7. Bishop, C.M. (2013) Model-based machine learning. Philos. Trans. A Math. Phys. Eng. Sci. 371
8. Lee, J.-G. et al. (2017) Deep learning in medical imaging: general overview. Korean J. Radiol. 18, 570–584
9. Guncar, 9 G. et al. (2018) An application of machine learning to haematological diagnosis. Sci. Rep. 8, 411
10. Koohy, H. (2017) The rise and fall of machine learning methods in biomedical research. F1000 Res. 6 http://dx.doi.org/10.12688/f1000research.13016.2
11. Young, J.D. et al. (2017) Unsupervised deep learning reveals prognostically relevant subtypes of glioblastoma. BMC Bioinf. 18, 381
12. Chen, H. et al. (2018) The rise of deep learning in drug discovery. Drug Discov. Today 23, 1241–1250
13. Grys, B.T. et al. (2017) Machine learning and computer vision approaches for phenotypic profiling. J. Cell Biol. 216, 65–71
14. Labovitz, D.L. et al. (2017) Using artificial intelligence to reduce the risk of nonadherence in patients on anticoagulation therapy. Stroke 48, 1416–1419
15. Mak K.-K., Pichika M.R. Artificial intelligence in drug development: present status and future prospects. Drug Discovery Today. 2019; 24:773–780.
16. Vyas M. Artificial intelligence: the beginning of a new era in pharmacy profession. Asian J. Pharm. 2018; 12:72–76.
17. Mak K.-K., Pichika M.R. Artificial intelligence in drug development: present status and future prospects. Drug Discovery Today. 2019; 24:773–780.
18. Sellwood M.A. Artificial intelligence in drug discovery. Fut. Sci. 2018; 10:2025–2028.
19. Zhu H. Big data and artificial intelligence modeling for drug discovery. Annu. Rev. Pharmacol. Toxicol. 2020; 60:573–589
20. Ciallella H.L., Zhu H. Advancing computational toxicology in the big data era by artificial intelligence: data-driven and mechanism-driven modeling for chemical toxicity. Chem. Res. Toxicol. 2019; 32:536–547.
21. Chan H.S. Advancing drug discovery via artificial intelligence. Trends Pharmacol. Sci. 2019; 40(8):592–604
22. Brown N. Royal Society of Chemistry; 2015. Silico Medicinal Chemistry: Computational Methods to Support Drug Design
23. Pereira J.C. Boosting docking-based virtual screening with deep learning. J. Chem. Inf. Model. 2016; 56:2495–2506.
24. Firth N.C. MOARF, an integrated workflow for multiobjective optimization: implementation, synthesis, and biological evaluation. J. Chem. Inf. Model. 2015; 55:1169–1180.
25. Zhang L. From machine learning to deep learning: progress in machine intelligence for rational drug discovery. Drug Discovery Today. 2017; 22:1680–1685.
26. Jain N. In silico de novo design of novel NNRTIs: a bio-molecular modelling approach. RSC Adv. 2015; 5:14814–14827.
27. Wang Y. A comparative study of family-specific protein–ligand complex affinity prediction based on random forest approach. J. Comput.-Aided Mol. Des. 2015; 29:349–360
28. King R.D. Comparison of artificial intelligence methods for modeling pharmaceutical QSARS. Appl. Artif. Intell. 1995; 9:213–233.
29. Álvarez-Machancoses Ó, Fernández-Martínez J.L. Using artificial intelligence methods to speed up drug discovery. Expert Opin. Drug Discovery. 2019; 14:769–777.
30. Fleming N. How artificial intelligence is changing drug discovery. Nature. 2018; 557 S55–S55.
31. Álvarez-Machancoses Ó, Fernández-Martínez J.L. Using artificial intelligence methods to speed up drug discovery. Expert Opin. Drug Discovery. 2019; 14:769–777.
32. Dana D. Deep learning in drug discovery and medicine; scratching the surface. Molecules. 2018; 23:2384.
33. Zang Q. In silico prediction of physicochemical properties of environmental chemicals using molecular fingerprints and machine learning. J. Chem. Inf. Model. 2017; 57:36–49.
34. Yang X. Concepts of artificial intelligence for computer-assisted drug discovery. Chem. Rev. 2019; 119:10520–10594.
35. Hessler G., Baringhaus K.-H. Artificial intelligence in drug design. Molecules. 2018; 23:2520.
36. Chai S. A grand product design model for crystallization solvent design. Comput. Chem. Eng. 2020; 135:106764.
37. Thafar M. Comparison study of computational prediction tools for drug–target binding affinities. Frontiers Chem. 2019; 7:1–19.
38. Öztürk H. DeepDTA: deep drug–target binding affinity prediction. Bioinformatics. 2018; 34:i821–i829.
39. Lounkine E. Large-scale prediction and testing of drug activity on side-effect targets. Nature. 2012; 486:361–367
40. Mayr A. DeepTox: toxicity prediction using deep learning. Frontiers Environ. Sci. 2016; 3:80.
41. Zhao YH, Abraham MH, Ibrahim A et al. Predicting penetration across the blood–brain barrier from simple descriptors and fragmentation schemes. J. Chem. Inf. Model. 47(1), 170–175 (2007)
42. Suenderhauf C, Hammann F, Huwyler J. Computational prediction of blood–brain barrier permeability using decision tree induction. Molecules 17(9), 10429–10445 (2012).
43. Guo M. A prototype intelligent hybrid system for hard gelatin capsule formulation development. Pharm. Technol. 2002; 6:44–52
44. Mehta C.H. Computational modeling for formulation design. Drug Discovery Today. 2019; 24:781–788.
45. Zhao C. Toward intelligent decision support for pharmaceutical product development. J. Pharm. Innovation. 2006; 1:23–35
46. Rantanen J., Khinast J. The future of pharmaceutical manufacturing sciences. J. Pharm. Sci. 2015; 104:3612–3638.
47. Hay M. Clinical development success rates for investigational drugs. Nat. Biotechnol. 2014; 32:40–51
48. Harrer S. Artificial intelligence for clinical trial design. Trends Pharmacol. Sci. 2019; 40:577–591.
49. Fogel D.B. Factors associated with clinical trials that fail and opportunities for improving the likelihood of success: a review. Contemp. Clin. Trials Commun. 2018; 11:156–164
50. Segler, M.H.S. et al. (2018) Generating focused molecule libraries for drug discovery with recurrent neural networks. ACS Cent. Sci. 4, 120–131
51. Huang, Z. et al. (2017) Data mining for biomedicine and healthcare. J. Healthc. Eng. 2017 http://dx.doi.org/10.1155/2017/7107629 Article ID 7107629, 2 pages
52. Zhang, Y. et al. (2017) Computer-aided clinical trial recruitment based on domainspecific language translation: a case study of retinopathy of prematurity. J. Healthc. Eng. 2017, 7862672
53. Mamoshina, P. et al. (2016) Applications of deep learning in biomedicine. Mol. Pharm. 13, 1445–1454
54. Seddon, G. et al. (2012) Drug design for ever, from hype to hope. J. Comput. Aided Mol. Des. 26, 137–150
55. Merk, D. et al. (2018) De novo design of bioactive small molecules by artificial intelligence. Mol. Inform. 37, 1700153
56. Klopman, G. et al. (2004) ESP: a method to predict toxicity and pharmacological properties of chemicals using multiple MCASE databases. J. Chem. Inf. Comput. Sci. 44, 704–715
57. Menden, M.P. et al. (2013) Machine learning prediction of cancer cell sensitivity to drugs based on genomic and chemical properties. PLoS One 8, e61318
58. Nascimento, A.C.A. et al. (2016) A multiple kernel learning algorithm for drugtarget interaction prediction. BMC Bioinf. 17, 46
59. Schneider, G. (2017) Automating drug discovery. Nat. Rev. Drug Discov. 17, 97–113
60. Matthews, H. et al. (2016) Omics-informed drug and biomarker discovery: opportunities, challenges and future perspectives. Proteomes 4 http://dx.doi.org/ 10.3390/proteomes4030028
61. Hamet, P. and Tremblay, J. (2017) Artificial intelligence in medicine. Metabolism 69, S36–S40
62. Hughes, J.P. et al. (2011) Principles of early drug discovery. Br. J. Pharmacol. 162, 1239–1249
63. Mohs, R.C. and Greig, N.H. (2017) Drug discovery and development: role of basic biological research. Alzheimers Dement. 3, 651–657
64. Katsila, T. et al. (2016) Computational approaches in target identification and drug discovery. Comput. Struct. Biotechnol. J. 14, 177–184
65. Wang, Q. et al. (2017) A novel framework for the identification of drug target proteins: combining stacked auto-encoders with a biased support vector machine. PLoS One 12, e0176486
66. Ferrero, E. et al. (2017) In silico prediction of novel therapeutic targets using genedisease association data. J. Transl. Med. 15, 182
67. Bakkar, N. et al. (2018) Artificial intelligence in neurodegenerative disease research: use of IBM Watson to identify additional RNA-binding proteins altered in amyotrophic lateral sclerosis. Acta Neuropathol. 135, 227–247
68. Reymond, J.-L. et al. (2010) Chemical space as a source for new drugs. Med. Chem. Commun. 1, 30–38
69. Okafo, G. et al. (2018) Adapting drug discovery to artificial intelligence. Drug Target Rev. 2018, 50–52
70. Segler, M.H.S. et al. (2018) Planning chemical syntheses with deep neural networks and symbolic AI. Nature 555, 604–610
71. Mayr, A. et al. (2016) DeepTox: toxicity prediction using deep learning. Front. Environ. Sci. 3, 80
72. Gayvert, K.M. et al. (2016) A data-driven approach to predicting successes and failures of clinical trials. Cell. Chem. Biol. 23, 1294–1301
73. Rodrigues, T. et al. (2018) Machine intelligence decrypts b-lapachone as an allosteric 5-lipoxygenase inhibitor. Chem. Sci. 9, 6899–6903
74. Luechtefeld, T. et al. (2018) Machine learning of toxicological big data enables readacross structure activity relationships (RASAR) outperforming animal test reproducibility. Toxicol. Sci. 165, 198–212
75. Bain, E.E. et al. (2017) Use of a novel artificial intelligence platform on mobile devices to assess dosing compliance in a Phase 2 clinical trial in subjects with schizophrenia. JMIR Mhealth Uhealth 5, e18
76. Perez-Gracia, J.L. et al. (2017) Strategies to design clinical studies to identify predictive biomarkers in cancer research. Cancer Treat. Rev. 53, 79–97
77. Deliberato, R.O. et al. (2017) Clinical note creation, binning, and artificial intelligence. JMIR Med. Inf. 5, e24
78. Corsello, S.M. et al. (2017) The Drug Repurposing Hub: a next-generation drug library and information resource. Nat. Med. 23, 405–408
79. 59 Hernandez, J.J. et al. (2017) Giving drugs a second chance: overcoming regulatory and financial hurdles in repurposing approved drugs as cancer therapeutics. Front. Oncol. 7, 273
80. Lozano-Diez, A. et al. (2017) An analysis of the influence of deep neural network (DNN) topology in bottleneck feature based language recognition. PLoS One 12, e0182580
81. Aliper, A. et al. (2016) Deep learning applications for predicting pharmacological properties of drugs and drug repurposing using transcriptomic data. Mol. Pharm. 13, 2524–2530
82. Galbusera, F. et al. (2018) Exploring the potential of generative adversarial networks for synthesizing radiological images of the spine to be used in in silico trials. Front. Bioeng. Biotechnol. 6, 53
Received on 10.11.2022 Modified on 21.11.2022
Accepted on 08.12.2022 ©AJRC All right reserved
Asian J. Research Chem. 2023; 16(1):83-90.
DOI: 10.52711/0974-4150.2023.00014