INTRODUCTION:

The COVID-19 pandemic continues to have a disastrous impact on the global population's health and well-being. Since the early days of this tragic disaster, there has been a rise in the investigation and usage of artificial intelligence (AI) and other data analytic techniques in a variety of domains. AI and machine learning (ML) have shown to be extremely effective in a variety of medical sectors, as well as playing a critical part in complex therapeutic scenarios. These systems have demonstrated a high level of accuracy in different applications, such as breast cancer, skin lesion classification, lung disease classification, Alzheimer, identifying diabetic retinopathy, and improving reconstruction for magnetic resonance imaging (MRI), X-ray and computerized tomography (CT) imaging^1,².

Conversely, ML analysis of genetic variants from asymptomatic, mild or severe COVID-19 patients is performed to classify, predict, and treat people based on their vulnerability or resistance to potential COVID-19 infection and their response to certain medications and cures^{3, 4}. These algorithms are trained to find patterns and features in huge amounts of data in order to make decisions and predictions based on new data. Eventually, these algorithms are used to suggest optimum combinations of drugs and provide best cure. Actually, a number of research labs and data centers have indicated that they are recruiting AI to search for treatments and vaccines^5,6. In this paper, we provide an extensive study on the role of AI and ML in delivering efficient responses towards COVID-19 pandemic, more specifically on the important practices in the domain of treatment and drug discovery. At the time of writing this manuscript, international health communities have proposed different general workflows and procedures for treatment and cure. Also, few vaccinations have been developed by some leading companies and approved by the World Health Organization (WHO); yet, no explicit drug was announced that can directly target the virus. Beside that, the virus continues to develop its mutations. Hence, we aim at pointing out the important directions in the field of drug discovery in order to pave possible paths for similar predicted viruses and future pandemics, beside COVID-19 and its mutations.

The availability of vast amounts of data, as well as the development and widespread availability of advanced computer systems capable of processing all of that data faster and more accurately than people, have enabled a rise in AI applications where various related disciplines have evolved. Such domains include machine learning, deep learning, and natural language processing. The rest of the paper is organized as follows. Section 2 presents a brief idea about the concepts of artificial intelligence, machine learning and deep learning. Section 3 covers COVID-19 origin and the genome structure. Drug discovery approaches are discussed in Section 4. Section 5 provides a discussion of the topic. Finally, the paper is brought to a conclusion in Section 6.

ARTIFICIAL INTELLIGENCE, MACHINE LEARNING AND DEEP LEARNING:

AI focuses on the development of intelligent systems that can learn from data and make appropriate judgments and predictions. ML is a subclass of AI that depends on the characteristic features and can learn based on experience without being explicitly programmed. Some of the leading ML approaches include random forest (RF), decision tree (DT), support vector machine (SVM), k-means, hierarchical clustering, and artificial neural networks (ANN). Moreover, deep learning (DL) is the subset of ML that can solve complex schemes through representation learning. Some of the notable DL techniques include convolutional neural network (CNN), recurrent neural network (RNN), and long short-term memory (LSTM)^{7, 8}.

ML algorithms attempt to emulate biological neural networks based on the mathematical structure to address complicated data-oriented problems. DL comprises innovative ANN-based ML algorithms that use multiple layers of processing units to extract higher-level characteristics from the data. Most of the supervised ML methods can operate with small, organized and labeled datasets, while DL techniques can work with raw, unstructured data and require significantly higher volumes. In healthcare and biomedical research, ML is used for a number of tasks including disease management, image analysis, and medical equipment. For example, Watson for Oncology, a program developed by IBM, has successfully forecasted medications for the treatment of cancer patients. Similarly, the Hanover Project at Microsoft suggests a customized cancer treatment^{9, 10}.The adoption of such developing technologies has the potential to considerably reduce the main challenge connected with medication repurposing and drug-disease correlation identification related to the current virus mutations and future pandemics.

COVID-19 ORIGIN:

The Coronavirus family is categorized into seven categories: Human Coronavirus OC43, Human Coronavirus 229e, Human Coronavirus NL63, SARS-CoV, Middle East Respiratory Syndrome Coronavirus (MERS-CoV), Human Coronavirus HKU1, and Wuhan Coronavirus. This novel coronavirus is the seventh type (COVID-19)^{11, 12}. Coronaviruses (CoV) are classified into four genera, including α−/β−/γ−/δ-CoV. The SARS-CoV-2 is a β-coronavirus, which is enveloped non-segmented positive-sense RNA virus (subgenus sarbecovirus, Orthocoronavirinae subfamily)¹³. α- and β-CoV are able to infect mammals, while γ- and δ-CoV tend to infect birds. Previously, six CoVs have been identified as human-susceptible virus, among which α-CoVs HCoV-229E, HCoV-NL63, β-CoVs HCoV-HKU1 and HCoV-OC43 with low pathogenicity that can cause mild respiratory symptoms similar to a common cold. The other two known β-CoVs, SARS-CoV and MERS-CoV lead to severe and potentially fatal respiratory tract infections¹⁴. The cross-sectional view of SARS-CoV-2 is shown in (figure 1). It consists of spike protein (S), hemagglutinin-esterase dimer (HE), nucleocapsid protein (N), an envelope protein (E), membrane glycoprotein/matrix (M), and single-strand RNA¹⁵.

The genomic sequence of SARS-CoV-2 was discovered to be 96.2% identical to a bat CoV RaTG13, and 79.5% identical to SARS-CoV. Based on virus genome sequencing, bats are suspected of being the virus's native host, and SARSCoV-2 could be transmitted from bats to people via unknown intermediate hosts. It is clear now that SARS-CoV-2 could use angiotensin-converting enzyme 2 (ACE2), the same receptor as SARS-CoV, to infect humans¹⁶.

Figure 1: Cross-sectional View of SARS-CoV-2 ¹⁵.

Genome Structure and Machine Learning:

The genome of CoVs has been discovered to include a variable number (6–11) of open reading frames (ORFs)¹⁷. Two-thirds of viral RNA, mainly located in the first ORF (ORF1a/b) translates two polyproteins, pp1a and pp1ab, and encodes 16 non-structural proteins (NSP), while the remaining ORFs encode structural proteins and accessory. The rest part of virus genome encodes four crucial structural proteins, including small envelope (E) protein, matrix (M) protein, spike (S) glycoprotein, and nucleocapsid (N) protein, and also numerous accessory proteins that affect the host innate immune response¹⁸. At the protein level, there are no amino acid substitutions that transpired in NSP7, NSP13, matrix, envelope, or accessory proteins p6 and 8b, except in spike protein, NSP2, NSP3, underpinning subdomain (receptor-binding domain (RBD))¹⁹. Another study²⁰ revealed that SARS-CoV-2 infectious capability and differentiation mechanism are influenced by mutations in NSP2 and NSP3. In this regard, Zhang et al.²¹ investigated COVID-19 genotypes in patients from various regions and discovered that SARS-CoV-2 had been mutated in several individuals in China. Tang et al.²² performed population genetic analysis on 103 SARS-CoV-2 genomes and identified two distinct SARS-CoV-2 evolution types, S type (~ 30%) and L type (~ 70%). L type strains, which are descended from S type strains, have evolved to be more contagious and aggressive¹⁸.

In fact, ML algorithms may be used to compare the viral genome to known genomes and find commonalities. The authors in ²³ have proposed the use of RF algorithm to identify the hosts for the influenza-A virus. Another example of such an approach is given in ²⁴. Such models can be extended to include the SARS-CoV-2 and its mutations. Authors in ²⁵ examined genomic mutations in the coding regions of SARS-CoV-2 and how they would affect protein secondary structure and solvent accessibility. According to the predictions, the mutation D614G in the viral spike protein, which has acquired a lot of attention, is unlikely to modify the protein's secondary structure or relative solvent accessibility. A dataset of point mutations was produced based on 6324 viral genome sequences, which can assist in the analysis of SARS-CoV-2 in various ways, particularly in tracing the virus's evolution. There were four different types of mutations found: insertion, deletion, synonymous, and nonsynonymous. The results of the investigation also revealed that the coding genes ORF6, ORF7a, ORF7b, ORF10, E, and M are the most stable, making them potentially appropriate for vaccine and drug development.

DRUG DISCOVERY APPROACHES:

Another wave of COVID-19 infections is sweeping the globe. Daily rates of new infections have risen dramatically in recent months around the world, and deaths have risen as well— this horrifying emergency is once again putting strain on the world's already overburdened healthcare system²⁶. To combat the pandemic and the resulting load on healthcare, scientists are putting AI solutions to the test. In the fight against infectious diseases, computational approaches have proven to be particularly useful in fundamental research, diagnosis, and therapy. In vaccine development, medication repurposing, and drug discovery, AI-based techniques have also emerged as a useful tool.

In this context, two well-known approaches to drug discovery are antibody development and small molecule development. During antibody formation, antibodies bind to the virus surface protein and cease binding to a host cell receptor. In small molecular development, computational tools are used to generate new molecules that function as ligands to obstruct target proteins²⁷. One of the major challenges in drug discovery is the massive search space for novel molecules²⁸. Scientists select and examine molecules that bind to biological targets that inhibit the replication of bacteria or viruses. The high-throughput screening experiments become very costly. As such, it is preferable to use computational tools to reduce the wide search space to mitigate these restrictions. In this regard, molecular docking and DL are two computational approaches for drug discovery that have received considerable attention. Molecular docking investigates how well a drug molecule attaches to biological targets using a three-dimensional simulation in which drug molecules (ligands) locate their position within the target (protein) site. However, there are two fundamental disadvantages to this approach: first, obtaining the 3D structure of the target protein is challenging; second, big simulations are costly and time-consuming. DL approaches, on the other hand, are attractive since they drastically reduce the time and cost of large-scale testing of candidate compounds and can provide automated design and screening of candidate molecules with some desired properties²⁷. Detailed investigations are provided in the following subsections.

Machine Learning Methods:

The massive amount of biomedical information has prompted the use of ML in numerous fields of biomedicine and the pharmaceutical business. ML's superior ability to extract higher-level features makes it a viable candidate for analyzing, binding, and interpreting heterogeneous biomedical data, and hence, to develop effective vaccinations and medical therapies for COVID-19 and its associated mutations.

Lee et al.²⁹ described a number of strategies for integrating AI and simulation-based approaches into the design of computational infrastructure. They demonstrated how a scalable computing infrastructure may enable novel methodologies on a variety of disparate and heterogeneous platforms. The authors explained how, while "In silico" drug discovery is a promising strategy, it is computationally intensive and difficult. They emphasized that there is an urgent need to improve "In silico" approaches in order to speed and select better lead compounds for further development in the drug discovery process. A basic difficulty in "In silico" is the requirement to span many timelines while traversing a vast combinatorial and chemical space. No one methodological technique can reach the needed accuracy of lead chemical selection while maintaining the required computing efficiency in this complicated landscape. In addition, Randhawa et al.³⁰ investigated the taxonomy of COVID-19 through the use of genetic markers and a decision tree methodology. The alignment-free method is a computationally efficient method for rapidly classifying new pathogens using only raw DNA sequence data. Using genomic sequences, conventional clustering approaches such as hierarchical clustering can be used to determine the virus's source. Alternately, a fuzzy logic system can be used to anticipate secondary structures of proteins based on the quantitative features of the amino acids utilized to encode the twenty most prevalent amino acids³¹. A combination of Least Absolute Shrinkage and Selection Operator (LASSO) and principal component analysis (PCA) can be used to analyze data on single-nucleotide polymorphism genetic variation in a supervised manner³².

The authors of ³³ suggested a data-driven strategy for drug repurposing using a combination of ML and statistical analysis tools. The authors began with a list of 6225 prospective medications, which were then whittled down through a series of processes that included a network-based knowledge mining technique and a connectivity map analysis method. Authors in ³⁴ discussed an interesting case of drug repurposing in which a Siamese neural network was used to identify the COVID-19 protein structure in comparison to Ebola and HIV-1 viruses. The benefit of this work is that the suggested model may be used directly with existing biological databases rather than with publicly available databases. Similarly, the authors in ³⁵ presented a method for discovering candidate compounds that may bind to the COVID-19 protein targets, which could then be utilized to produce candidate treatment drugs.

In ³⁶, the capacities of ML in creating, analyzing, and monitoring the delivery of pharmaceuticals were investigated. It was demonstrated that ML may aid pharmaceutical makers in comprehending and utilizing systems to operate their organizations more effectively, and then provide them with the flexibility to adjust to shifting situations and needs for swift product delivery.

Deep Learning Algorithms:

DL algorithms have appeared in a variety of contexts and implementations across a wide range of COVID-19 applications, especially in treatment and drug discovery. In this regard, the authors of ³⁷ developed a bioinformatics technique for monitoring treatment that included Generative Adversarial Networks (GANs), Extreme Learning Machines (ELMs), and LSTM. This strategy combines various components of information from a variety of structured and unstructured data sources to create platforms that can support in treatment monitoring. The technique is based on RNNs with a large number of inputs. RNNs' prediction accuracy is influenced by their ability to recall previous events prior to learning the underlying relationship of the data in order to reach the hidden layers. As stated in ³⁷, coronavirus infection can result in vascular inflammation, arrhythmias, and myocarditis. They presented a methodology for minimizing the risk of cardiovascular problems and ensuring appropriate responses to various treatment modalities. Whereas GANs are trained on the data distribution of interest, the discriminative network distinguishes between GAN-generated candidate datasets and real-world data. GANs can be used in a variety of ways, as demonstrated in ³⁸. According to the findings of this work, coronaviruses that had previously eluded detection by conventional methods can be visualized and detected using a combination of transmission electron microscopy and whole-genome sequencing of culture supernatant. Figure 2 depicts the proposed neural network model and the GAN.

Figure 2: Application of GAN for visualization and detection of coronavirus³⁸.

Predictions of protein structure can also be made using computational models. Genetically encoded amino acid sequences dictate the protein's 3D structure, which in turn influences its function. Both template modeling and template-free modeling can be used for protein structure prediction; however the latter is more commonly used. AlphaFold, developed by Senior et al.³⁹, is a technique for predicting protein structure in the absence of a known related structure. For estimating the difference between amino acid residues, the AlphaFold model makes use of features collected from comparable amino acid sequences via multiple sequence alignment, which is a dilated version of ResNet. A "potential of mean force" is utilized to define the protein's structure based on these assumptions. According to Segler et al.⁴⁰, they developed a three-part neural network pipeline, which was combined with a Monte Carlo Tree Search approach, to mine a structured database of chemical reactions in order to recognize how different compounds are shaped hierarchically from reactions between simpler compounds. Fauquer et al.⁴¹ presented an approach for mining an unstructured database to uncover stylized associations between gene-disease pairs based on the information included in the database. Similar approaches to constructing a knowledge graph connecting human proteins, viral proteins, and medications were described by Ge et al.⁴², who used databases to record the interactions between these items. The graph is used to estimate whether or not a candidate medicine will be effective.

Additionally, several researches sought to anticipate the binding affinities of proteins to ligands in order to address the medication repurposing problem. Ligands are tiny chemicals that interact with proteins to initiate a signaling cascade, which can be either activating or inhibiting. Hu et al.⁴³ predicted these affinities using a multitask neural network, identifying a list of eight SARS-CoV-2 related proteins that they seek to target using a list of 4895 drugs. They also assessed the particular areas of every target protein where binding is anticipated to occur in an effort to improve model interpretability. Zhang et al.⁴⁴ identified potential inhibitors of the 3C-like protease using a fully connected neural network architecture trained to forecast binding affinities using the PDBbind database. They created a template model of the target protein using its SARS version and searched databases of known compounds as well as tripeptides for potential treatments that aim this protein.

Amilpur and Bhukya²⁷ suggested a strategy for developing novel compounds that are capable of binding to the COVID-19 protease based on RNNs and LSTM models. They investigated 3CLPro, the primary protease, as a therapeutic target and illustrated the use of docking simulations to assess target structure binding affinities. They collected almost 2.9 million compounds initially from two well-known molecular databases, Moses and ChemBL. They recovered over 2.5 million molecules during a cleaning process. These molecules are fed into the proposed model as one-hot encoded input vectors. The model generates a multinomial probability distribution and learns how realistic and plausible drug-like compounds are. After that, it generates new drug candidates by sampling from the distribution. By employing generative models for molecule generation, the search field is narrowed. The authors discovered that 80% of the compounds synthesized have a docking free energy of less than 5.8 kcal/mol. With a docking score of 8.5 kcal/mol, the top-generated candidate had the maximum binding affinity. The model architecture and molecular generation process is demonstrated in (figure 3).

Figure 3: Model architecture and molecular generation process²⁷.

Bioinformatics and Imagery Systems:

Using 3D pictures of lung postmortem tissues taken from COVID-19 patients, Li et al.⁴⁵ provided novel insights into disease processes contributing to mortality that could impact front-line treatment decisions, such as a virtual histology of cubic-millimeter volumes of diseased lung. Using fluorescence labeling and tissue clearing techniques, emerging methodologies for high-resolution, massively multiscale imaging of tissue microstructure allow the recording of 3D histology, which may provide new understandings into the nature of SARS-Cov-2 infection and COVID-19 disease processes. Fully three-dimensional study of the microstructure of tissue samples can assist disambiguate such sectioning errors and may also help improve pathologic interpretation. In this context, the value of 3D imaging is that it enables considerably finer tissue sampling than typical 2D slide-based histology, which may offer extra information to enhance standard histology.

Furthermore, the authors in ⁴⁶ modelled the COVID-19 pharmaceutical compound by adjusting the percentage of chemicals in medicinal ingredients using meta-heuristic procedures, which are highly sophisticated and difficult to compute mathematically. To identify the optimal percentages of components in the COVID-19 pharmaceutical compound, a relationship between the frontend and backend was created. WebHDFS and

Spark API are the standard tools of Apache's backend, which is integrated with the Deep AI Core Engine System that can be accessed and run by a console in the terminal or on the Web. For the development of the Core Engine System, several different technologies were employed, from data extraction to processing to the transmission of the findings. Another technology employed by the authors to perform computing and analysis on massive datasets is the MapReduce technology.

On the basis of two different perspectives, Kabra and Singh⁴⁷ developed an algorithmic therapy that uses viral sequences to identify peptides with activity against various strains of the COVID-19 virus. These peptides were then examined for effectiveness against proteases of COVID-19 and tested for cross protection against various virus strains. The authors used computational methods to predict a library of these peptides to further identify peptides from diverse molecular sequences, as the recognition and design of these peptides is a resource-intensive methodology. Virus replication and transcription require the primary protease (M^pro) found in SARS-CoV-2, hence the researchers decided to go after this enzyme. Using the technique, four peptides were shown to have substantial binding activity and the ability to block protease action. Docking of these selected peptides along with their interactions against the M^pro and their docked poses is shown in (figure 4).

DISCUSSION:

In the near future, AI is likely to play an increasingly prominent role in all sectors of healthcare. However, putting such concepts into action in the real world is riddled with obstacles and has significant limitations. To give an example, the capacity of ML and DL algorithms to be effective over a wide range of inputs and applications is critical to their success. When applied to a diversified group of datasets, restricted context models always run the risk of failing at the broadest level of analysis. A reliable and balanced strategy to developing and governing AI techniques is essential due to the wide range of applications for which AI is being used.

In the field of computational biology and medicine, AI has been utilized to partially understand COVID-19 and to develop novel therapeutic compounds that are effective against the virus. Given that these are still preliminary findings, there is a tremendous mandate for AI research in this field, for example, to analyze the genetics and chemistry of the virus and advise strategies to rapidly produce vaccinations and treatment medications for its mutations. Also, it may be true that the experimental evidence in many studies is offered based on hundreds or thousands of observations, however, it is necessary to apply these techniques in real-world scenarios, where databases contain higher volumes of cases, each with a wide range of variability. This has become a significant issue in determining acceptable techniques to handle COVID-19 concerns. Further, because the coronavirus may experience more natural selection and change in a geographic manner, it is critical to understand this evolution and the uncertainty that may be associated with it. In order to successfully address this issue, it is necessary to quantify the degree of ambiguity in current knowledge about the condition in order to be able to evaluate the risks and advantages of the clinical and social policies that are being implemented.

CONCLUSION:

The purpose of this research is to present a detailed investigation of how the AI performed in domains connected to COVID-19, notably in the areas of drug discovery search. Massive amounts of data are being generated as a result of the extraordinary shift away from symptom-based treatment and toward molecular and cellular-based therapy. In this scenario, AI can add value by accelerating the pace of biomedical discovery and therapy availability. ML methods and DL algorithms have appeared and been implemented in a variety of COVID-19 applications, particularly in treatment and drug discovery. Moreover, bioinformatics approaches and imagery systems have the potential to provide unique insights into disease processes that could influence treatment decisions and aid in the adjustment of the percentage of chemicals in pharmaceutical substances, among other things.

In order to develop predictive therapeutic strategies against similar future pandemics, future AI research should benefit from a standardized collection of patient data. Our belief is that AI techniques can augment the ability to change and adapt current models, as well as combine them with primary clinical knowledge, to meet present challenges as well as novel virus strains or mutations. The goal is to enhance the adoption of highly precise and effective AI technology to combat future pandemics.

ACKNOWLEDGEMENTS:

The author would like to express his gratitude and grateful appreciation to the Kuwait Foundation for the Advancement of Sciences (KFAS) for financially supporting this project. The project was fully funded by KFAS under project code: PN20-13NH-03.

CONFLICTS OF INTEREST:

The author declares no conflict of interest.

REFERENCES:

1. D. P. Fan et al. Inf-Net: Automatic COVID-19 Lung Infection Segmentation From CT Images. IEEE Transactions on Medical Imaging. 2020; 39(8): 2626-2637doi: 10.1109/TMI.2020.2996645

2. R. M. Balajee, K. Venkatesh. A Survey on Machine Learning Algorithms and finding the best out there for the considered seven Medical Data Sets Scenario. Research J. Pharm. and Tech. 2019; 12(6):3059-3062.

3. Manisha Rokade, Pradnya Khandagale. Coronavirus Disease: A Review of a New Threat to Public Health. Asian J. Pharm. Res. 2020; 10(3):241-244.

4. Ahmad Alimadadi, Sachin Aryal, Ishan Manandhar, X Patricia B. Munroe, Bina Joe, and X Xi Cheng. AI and Machine Learning for Understanding Biological Processes - Artificial intelligence and machine learning to fight COVID-19. Physiol Genomics. 2020; 52: 200–202 doi:10.1152/physiolgenomics.00029.2020.

5. Browning, L., Colling, R., Rakha, E., Rajpoot, N., Rittscher, J., James, J. A., and Verrill, C. Digital pathology and artificial intelligence will be key to supporting clinical and academic cellular pathology through COVID-19 and future crises: the PathLAKE consortium perspective. Journal of Clinical Pathology. 2021; 74(7): 443-447.

6. K. Nalini Devi, Juliana Shanthi Rosy. ECMO therapy for COVID-19. Asian J. Nursing Education and Research. 2021; 11(2): 299-301.

7. Yamuna M, Elakkiya A. Mathematical Models in Drug Discovery, Development and Treatment of Various Diseases – A Case Study. Research J. Pharm. and Tech. 2017; 10(12): 4397-4401.

8. Xin, Y., Kong, L., Liu, Z., Chen, Y., Li, Y., Zhu, H., and Wang, C. Machine learning and deep learning methods for cybersecurity. IEEE ACCESS. 2018; 6: 35365-35381.

9. Anish Chittora, Mary Mekala A. Discovery of Drug and Medicine Using Data Mining Techniques. Research J. Pharm. and Tech. 2017; 10(12): 4147-4151.

10. Johnson, K. B., Wei, W. Q., Weeraratne, D., Frisse, M. E., Misulis, K., Rhee, K. and Snowdon, J. L. Precision medicine, AI, and the future of personalized health care. Clinical and Translational Science. 2021; 14(1): 86-93.

11. Rajesh Kumar, Abdullah Aman Khan, Sinmin Zhang, WenYong Wang, Yousif Abuidris, Waqas Amin, and Jay Kumar, “Blockchain-Federated-Learning and Deep Learning Models for COVID-19 detection using CT Imaging. Journal of Latex Class Files. 2020; 14(8).

12. Paresh A. Patil, Ritik. S. Jain. Theoretical Study and treatment of Novel COVID-19. Res. J. Pharmacology and Pharmacodynamics. 2020; 12(2): 71-72.

13. Zhu N, Zhang D, Wang W, Li X, Yang B, Song J, et al. A novel coronavirus from patients with pneumonia in China, 2019. N Engl J Med. 2020; 382(8): 727–33

14. Yin Y, Wunderink RG. MERS, SARS and other coronaviruses as causes of pneumonia. Respirology. 2018; 23(2):130–7

15. Su, Z., Bentley, B., & Shi, F. (2020). Articial Intelligence-based Disease Surveillance Amid COVID-19 and Beyond. A Systematic Review Protocol. 1–15. https://doi.org/10.21203/rs.3.rs-74191/v1.

16. Zhou P, Yang XL, Wang XG, Hu B, Zhang L, Zhang W, et al. A pneumonia outbreak associated with a new coronavirus of probable bat origin. Nature. 2020. https://doi.org/10.1038/s41586-020-2012-7

17. Song Z, Xu Y, Bao L, Zhang L, Yu P, Qu Y, et al. From SARS to MERS, thrusting coronaviruses into the spotlight. Viruses. 2019; 11(1): E59. https://doi.org/10.3390/v11010059

18. Yan-Rong Guo1, Qing-Dong Cao, Zhong-Si Hong, Yuan-Yang Tan, Shou-Deng Chen, Hong-Jun Jin, Kai-Sen Tan, De-Yun Wang and Yan Yan. The origin, transmission and clinical therapies on coronavirus disease 2019 (COVID-19) outbreak – an update on the status. Military Medical Research. 2020; 7:11, https://doi.org/10.1186/s40779-020-00240-0

19. Wu A, Peng Y, Huang B, Ding X, Wang X, Niu P, et al. Genome composition and divergence of the novel coronavirus (2019-nCoV) originating in China. Cell Host Microbe. 2020. https://doi.org/10.1016/j.chom.2020.02.001

20. Angeletti S, Benvenuto D, Bianchi M, Giovanetti M, Pascarella S, Ciccozzi M. COVID-2019: the role of the nsp2 and nsp3 in its pathogenesis. J Med Virol. 2020. https://doi.org/10.1002/jmv.25719

21. Zhang L, Shen FM, Chen F, Lin Z. Origin and evolution of the 2019 novel coronavirus. Clin Infect Dis. 2020. https://doi.org/10.1093/cid/ciaa112

22. Tang X, Wu C, Li X, Song Y, Yao X, Wu X, et al. On the origin and continuing evolution of SARS-CoV-2. Natl Sci Rev. 2020. https://doi.org/10.1093/nsr/nwaa036

23. A. Panday, ``Virus outbreak drives down automobile sales in March,'' Livemint, Apr. 2020. [Online]. Available: https://www.livemint.com/companies/news/automakers-hit-hard-bycovid-19%-march-sales-decline-over-50-11585749547879.html

24. A. Panday, ``Amid Covid-19 crisis, commercial vehicle sales crash 89% in March,'' Livemint, Apr. 2020. [Online]. Available: https://www.livemint.com/companies/news/amid-covid-19-crisiscommercial%-vehicle-sales-crash-89-in-march-11585756974053.html

25. Thanh Thi Nguyen, Pubudu N. Pathirana, Thin Nguyen, Quoc Viet Hung Nguyen, Asim Bhatti, Dinh C. Nguyen, Dung Tien Nguyen, Ngoc Duy Nguyen, Douglas Creighton, and Mohamed Abdelrazek, “Genomic Mutations and Changes in Protein Secondary Structure and Solvent Accessibility of SARS-CoV-2 (COVID-19 Virus),” bioRxiv preprint doi: https://doi.org/10.1101/2020.07.10.171769; 2021

26. Centers for Disease Control and Prevention. (2022, Feb 1). Omicron Variant: What You Need to Know. https://www.cdc.gov/coronavirus/2019-ncov/variants/omicron-variant.html

27. Amilpur, S., & Bhukya, R. (2022). Predicting novel drug candidates against Covid-19 using generative deep neural networks. Journal of Molecular Graphics and Modelling. 2022; 110: 108045.

28. J.-L. Reymond, L. Ruddigkeit, L. Blum, R. van Deursen, The enumeration of chemical space, Wiley Interdiscipl. Rev.: Comput. Mol. Sci. 2 (5) (2012) 717–733.

29. Hyungro Lee et al., “Scalable HPC & AI Infrastructure for COVID-19 Therapeutics,” arXiv:2010.10517v1 [cs.DC] 20 Oct 2020

30. Randhawa, G. S., Soltysiak, M. P., El Roz, H., de Souza, C. P., Hill, K. A., and Kari, L. Machine learning using intrinsic genomic signatures for rapid classification of novel pathogens: COVID-19 case study. bioRxiv, doi: https://doi.org/10.1101/2020.02.03.932350

31. Nguyen, T., Khosravi, A., Creighton, D., and Nahavandi, S. Multi-output interval type-2 fuzzy logic system for protein secondary structure prediction. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems. 2015; 23(5): 735-760

32. Araghi, S., and Nguyen, T. T. A hybrid supervised approach to human population identification using genomics data. IEEE/ACM Transactions on Computational Biology and Bioinformatics. 2019. doi: 10.1109/TCBB.2019.2919501

33. P. Li, X. Jiang and H. Shatkay. Figure and caption extraction from biomedical documents. Bioinformatics. 2019; 35: 4381-4388.

34. N. Siegel, N. Lourie, R. Power and W. Ammar, "Extracting scientific figures with distantly supervised neural networks", Proc. 18th ACM/IEEE Joint Conf. Digit. Libraries, pp. 223-232, 2018

35. D. C. Comeau, C.-H. Wei, R. I. Doğan and Z. Lu. PMC text mining subset in BioC: About three million full-text articles and growing. Bioinformatics. 2019; 35: 3533-3535

36. Kotta Kranthi Kumar. Importance and Applications of Artificial Intelligence (Metastorm Software) in Pharmaceutical Process Life-Cycle. Res. J. Pharma. Dosage Forms and Tech. 2019; 11(2): 116-120.

37. Mohammad (Behdad) Jamshidi et al. Artificial Intelligence and COVID-19: Deep Learning Approaches for Diagnosis and Treatment,” IEEE Special Section On Emerging Deep Learning Theories And Methods For Biomedical Engineering. June 24, 2020.

38. N. Zhu, D. Zhang, W. Wang, X. Li, B. Yang, J. Song, X. Zhao, B. Huang, W. Shi, R. Lu, and P. Niu. China novel coronavirus investigating and research team. A novel coronavirus from patients with pneumonia in China, 2019. New England J. Med. 2020; 382(8): 727733

39. Senior AW, Evans R, Jumper J, Kirkpatrick J, Sifre L, Green T, et al. Improved Protein Structure Prediction Using Potentials from Deep Learning. Nature. 2020; 577(7792): 706-710. doi:10.1038/s41586-019-1923-7

40. Segler MH, Preuss M, Waller MP. Planning Chemical Syntheses with Deep Neural Networks and Symbolic AI. Nature. 2018; 555(7698): 604-610.

41. Fauqueur J, Thillaisundaram A, Togia T. Constructing Large Scale Biomedical Knowledge Bases from Scratch with Rapid Annotation of Interpretable Patterns. In: Proceedings of the 18th BioNLP Workshop and Shared Task. 2019: 142-151

42. Ge Y, Tian T, Huang S, Wan F, Li J, Li S, et al. A Data-Driven Drug Repositioning Framework Discovered a Potential Therapeutic Agent Targeting COVID-19. bioRxiv preprint bioRxiv:20200311986836. 2020

43. Hu F, Jiang J, Yin P. Prediction of Potential Commercially Inhibitors against SARS-CoV-2 by Multi-Task Deep Model. arXiv preprint arXiv:200300728. 2020

44. Zhang H, Saravanan KM, Yang Y, Hossain MT, Li J, Ren X, et al. Deep Learning Based Drug Screening for Novel Coronavirus 2019-nCov. Research Gate preprint 2020020061v1. 2020

45. Guang Li, Sharon E. Fox, Brian Summa, Bihe Hu, Carola Wenk, Aibek Akmatbekov, Jack L.Harbert, Richard S. Vander Heide, J. Quincy Brown. Multiscale 3-dimensional pathology findings of COVID-19 diseased lung using high-resolution cleared tissue microscopy. doi: https://doi.org/10.1101/2020.04.11.037473

46. Cholissodin et al. Smart Development of Big Data App for Determining the Modelling of Covid-19 Medicinal Compounds Using Deep AI Core Engine System. J. Phys. 2020; 1665: 1 doi:10.1088/1742-6596/1665/1/012003Conf. Ser. 1665 012003

47. Kabra, R., & Singh, S. Evolutionary artificial intelligence based peptide discoveries for effective Covid-19 therapeutics. Biochimica et Biophysica Acta (BBA)-Molecular Basis of Disease. 2021; 1867(1): 165978.

Received on 28.09.2022 Modified on 18.04.2023

Research J. Pharm. and Tech 2023; 16(11):5384-5391.

DOI: 10.52711/0974-360X.2023.00872