Artificial Intelligence-based strategies for drug repurposing
Sonakshi, Preeti P.1, Trapti Sharma1, Kamal Rawal#1
1. Amity Institute of Biotechnology, Amity University Uttar Pradesh, India.
#Corresponding Author
Email ID: kamal.rawal@gmail.com
Centre for Computational Biology and Bioinformatics,
AIB Amity University, Noida
Keywords:
Abstract:
Traditional drug discovery is an old strategy that was used earlier when technology was nascent, and the field of medicine and drugs was not as advanced as today. Today drug repurposing is widely used to make drugs available for the treatment of difficult diseases and fast mutating diseases. Existing and even rejected drugs are quickly repurposed to find effective medication. It can also be used for treating specific patients with personalized medicines. Drug repurposing is far more effective given the constraints of time in the discovery of drugs de-novo. In repurposing, there is no need to evaluate new molecular entities.
In drug repositioning, a combination of approaches are used from either computational to experimental to in silico-based Drug repurposing is today the way forward and a robust strategy utilizing available and historical medicines, that have already been tested as safe for use by humans to treat emerging and difficult diseases like COVID 19. In the case of COVID 19 modern technology such as AI and ML was effectively used to identify Remdesivir amongst other drugs to treat COVID 19.
1. Introduction
Drug repurposing is widely used today using existing and even rejected drugs to quickly develop new drugs to control a pandemic of finding an effective medication for serious illnesses. It can also be used for treating specific patients with personalized medicines. A broad process and targets of drug repurposing are represented in figure 1.1. Drug repurposing is far more effective given the constraints of time in the discovery of drugs de-novo. In drug repurposing, biological targets are the point of focus for determining the most suitable target for inhibiting drug development. Thus there is no need to evaluate new molecular entities.
In drug repositioning, a combination of approaches are used from either computational to experimental to in silico-based Drug repurposing is today the way forward and a robust strategy utilizing available and historical medicines, that have already been tested as safe for use by humans to treat emerging and difficult diseases like COVID 19.
Previously, we have developed several machine learning and bioinformatics platforms. These include, text mining and network biology-based systems [Jagannadham et al., 2016], vaccine discovery systems [Rawal et al. 2021], next-generation sequencing analysis systems for cancer and other genomes [Preeti et al.,2021, Rawal et al 2011, Mandal et al 2006].
1.1 SARS CoV-2
The SARS-Cov2 virus closely related to SARS (Severe Acute Respiratory Syndrome) of the Coronaviridae family emerged in Wuhan, China which is a virus ie highly the main derivative for COVID-19.
SARS-Cov2 is an enveloped positive-sense ssRNA beta coronavirus and has a genome of 30,000 nucleotides that code for 29 proteins i.e. the 4 structural proteins, 16 non-structural proteins, and 9 accessory proteins. COVID 19 pandemic has resulted in scientific research in antiviral drugs for COVID 19.
2. Drug Repurposing
2.1 Conventional V/s Drug Repositioning
There are two approaches to drug discovery as shown in Figures 2.1 & 2.2
Conventional drug development
The favourable outcome is very low (only 2.01%) for the development of a new molecular entity,
Time period involved is generally 10–15 years
Cost concerns due to high costs
Requires large investments
Drug repositioning
It involves the utilization of existing drugs for developing new therapeutics
The process of drug development is quicker,
Highly economical
Less risk
A repositioned drug bypasses the initial steps of traditionaltion drug discovery and goes to preclinical and clinical trials. This narrows down the risks besides lowering costs.
More than 80 original clinical trials are ongoing at present for COVD 19, including old drugs3 and potential new drugs. The recovery of discarded drugs & reusing old drugs helps increase patients’ lives. For a drug to be repurposed, for COVID 19 it requires the use of AI to assess biological and pharmaceutical aspects and interpret the mechanism of action of drugs.
3. AI approaches in Drug Repurposing
AI-based repurposing has transformed drug availability with quick discovery. Machine learning (ML)4 further enhances drug repurposing by generating new models that evaluate various data from a large data sample of patients or existing drugs. The deep learning (DL) technique of AI with its extraction feature enables the extraction of important information from a huge amount of data which is more result-oriented, predictive in comparison to another computer/computational models. Different DL algorithms that are being used for drug discovery for COVID 19 are artificial neural network (ANN), long short-term memory (LSTM) and, convolutional neural network (CNN). DL is the main approach for drug repurposing due to its extraction feature. Virtual screening of drugs that are repurposed & chemical entities new is also being used depicted in Figure 3.1. For VS, ML-based molecular docking is being used extensively for drug repurposing. ML4-based molecular docking requires:
dataset either of drug-like molecules or approved drugs.
target protein three-dimensional structure.
software for molecular docking.
The molecular docking helps identify chemical molecules that could bind to various CoV-2 proteins that may inhibit, slow down the viral replication and its growth.
SARS-CoV-2 structural proteins are targeted using small molecules in different databases to develop new & also for repurposing.
Through VS and molecular docking using ML/DL approach.
De-novo discovery of drug-using generative approaches viz variational autoencoders (VAE) and Generative Adversarial Networks (GAN) is feasible.
The generative models5 VAE and GAN generate sequences of atoms. This helps in creating unique drug molecules with greater diversity. The autoencoders direct the molecules into a vector that captures the element properties, its functional group, and bond order. For a drug-generative network design, the information about (a) collection of drug-like molecules, (b) a feature-representation of these molecules in silico, (c) method to increase the diversity of the molecules, and (d) screening and modification of the altered molecules, are considered.
AI, with strong computational6 power in computational biology and medicine, has helped to understand COVID-19 for drug discovery.
AI has helped in determining the protein structures of the virus, necessary for drug or vaccine discovery.
DL models can greatly assist better epitope prediction and in creating small molecule drugs thus reducing chances of failure in trials.
3.1 Deep Learning AI technique for Drug repurposing
The striking feature of DL is that it is its adaptability in the design of neural systems such as repetitive neural networks (RNN), convolutional neural networks (CNN), deep belief networks (DBN), and complete associated fully connected feedforward neural network (FNN)feed-forward systems. The AI drug repurposing strategy is represented in Figure 3.2.
Another network viz the Convolutional7 Neural Network ie CNN is found to be practically suitable for image processing. CNN uses filters or small matrices of weights that perform a convolution operation on local image patches which reduces the number of weights. CNN is widely used in the analysis of chemical images to obtain valuable inputs on drug therapeutic functions. For example, CNN has been effectively utilized for predicting the binding affinity of small molecules to proteins using AtomNet.
3.2 Graph Representational Learning
As has been mentioned above, network medicine is one way to repurpose drugs. In graph representational learning a graph is produced with interlinkages, b/w drugs, diseases, & proteins. This is then used to predict links between the approved drugs and diseases as was done for COVID-19 in the search for repurposed drugs. The graph method used for repurposing drugs is assuming high importance which makes it possible to depict edges and nodes as low dimensional vectors. Feature vectors of diseases and drugs and respective commonalities are measured which enables the identification of effective drugs for the particular disease. Graph methodology has a limitation of scalability. In reality with actual data, graphs will be very wieldy. In the medical knowledge graph, the data points may be several million.
To handle large-scale data the present ML systems viz PyTorch & TensorFlow are not sufficient for graphs of large scale. Hence, a system such as GraphVite is capable of handling large-scale data that is required for use in drug repurposing.
Knowledge graph BenevolentAI's is a large source of medical information, through machine learning. BenevolentAI was used as a drug repurposing tool. The drug used to treat rheumatoid arthritis viz baricitinib could be used as a drug to treat COVID-19, Baricitinib acts as an inhibitor of AP2-associated protein kinase 1 (encoded by AAK1).
Drug repurposing requires:
Repurposed Drug Database
Open Chemical Database
Drug Database is required input to the model.
Different algorithms could be applied to the input data to obtain the required drug.
3.3 Data Search Methodology
The database search8 for drug repurposing was done through the Dimensions database along with PRISMA guidelines. Dimension is a comprehensive database created by digital science with the help of information from over 100 leading research organizations. PRISMA Database has data from patents, publications, clinical trials, and funding agencies. Studies on drug discovery or drug repurposing using AI/ML approaches are also included. Table 1. Summary of findings for the development and repurposing drug in COVID-19 using AI and ML. Figure 3.3 depicts a flow diagram of PRISMA for drug development and repurposing drug for COVID-19 using AI/ML.
3.4 Power/Use of AI in Drug repurposing
AI has the ability to perform simulation tests for a potential drug by screening a large number (millions) of chemical compounds from different databases. The COVID 19 infection to management using AI is depicted diagrammatically in figure 3.4. It can also identify de novo drugs for COVID 19 targets, to reduce their infectivity. In the search for a suitable drug, the AI drug discovery company Imperial College London and BenevolentAI used their in-house9 developed algorithms to mine the data to find adaptor-associated protein kinase 1 (AAK1) that could be a possible target for COVID-19. Out of 378 known AAK1 inhibitors BenevolentAI assessed Baricitinib, an approved drug against rheumatoid arthritis as the best inhibitor. CVL218, discovered through a Natural Language Processing (NLP) AI model is a strong PARP1 (Poly (ADP-Ribose) Polymerase 1 inhibitor. This was possible by using the Biomedical Entity Relation Extraction (BERE) approach to the database from PubMed. Deep learning-based drug-target interaction model MT-DTI - Molecule Transformer-Drug Target Interaction deep learning drug target interaction model has been used to identify antiviral drugs for SARS-CoV-2 proteins10 (the 3CLpro, RdRP, endoRNAse, helicase, 2′-O-ribose methyltransferase, and 3′-to-5′ exonuclease,). SMILES ie ‘simplified molecular-input line-entry system’ strings and amino acid sequences have been used for 1D string inputs to target proteins without any 3D structures. Figure 3.4. Knowledge graph (KG) based DL method for drug repurposing in COVID-19 (CoV-KGE) has also been used. RotatE, a DL approach developed by Amazon, AWS-AI is used to develop a KG from DrugBank and a large number of PubMed publications (24 million). Using enrichment analysis of SARS-CoV-2 proteomics and data transcriptome plus input data of ongoing clinical trials, and drug-gene signatures 41 drugs were identified for repurposing. The drug-binding ability of SARS-CoV-2 3CLpro of potential protease inhibitors was evaluated through ML model SDBR (Structure-based drug repositioning). This was achieved through gradient-boosting11 decision tree (GBDT) utilizing a 2-D fingerprint-based DL on a large number (approx. 314) of SARS-CoV-2/SARS-CoV-3CLpro inhibitors. Of 8565 potential drugs evaluated from DrugBank, the top 20 drugs including off-market investigational drugs were selected as potent inhibitors. ChemAI,” (a deep ligand-based), a DL network model, screened approx 220 million data points across 3.6 million molecules from ZINC databases. Thus inhibitory potentials of approx. 900 compounds to the SARS-CoV-2 the papain-like protease (PLP) and 3CLpro were evaluated as depicted in Figure 3.5 below. From the DrugBank a list of 30,000 possible compounds was identified according to their closeness to known drugs, predicted inhibitory potentials, and toxicity, for drug repurposing. The top-ranked compounds are available as a library at https://github.com/ml-jku/sars-cov-inhibitors-chemai. A deep docking platform trained on a neural network to predict the results of docking simulation has also been used for predicting suitable drugs for COVID 19. From the ZINC database, a set of 3 million candidate 3CLpro inhibitors were identified which were subsequently narrowed down to 1000 compounds for drug repurposing after docking simulation. Figure 3.5. To predict protein-ligand binding affinities of viral proteins, a multitask neural network model against a database of 4895 drugs was exploited. 10 potential drugs with strong binding affinity to target proteins was identified. A Python-based DL toolkit, DeepPurpose based on an encoder-decoder framework is used as a case study on SARS-CoV-2 3CLpro with 13 potential repurposing candidates identified. A random forest algorithm on data from Smith and Smith, 2020 has also been used. The models as above were then superimposed on CureFFI & DrugCentral datasets containing 1495 and 3967 drugs, respectively. The model screened the compounds from BindingDB datasets and identified 19,000 drug candidates that could strongly bind to either the S-protein or the ACE2-S protein complex. Ligand Design a DL platform was used to identify existing drugs and experimental medicines with the potential to inhibit SARS-CoV-2 infection. The database used was PoylpharmDB 10,224 drugs along with a list of 8700 proteins to interact with the identified existing and experimental drugs. Cyclica’s MatchMaker TM technology was used to generate the interactions using, a DL model acting on the entire human proteome to predict the drug molecules binding to protein pockets.
The databases, ChemDiv, and TargetMol were used to find promising compounds to target 3CLpro protein. A systems biology and AI-based approach, viz the TPMS - Therapeutic Performance Mapping System technology was used to repurpose drugs and drug combinations for COVID-19. The Denovicon computational platform was used to perform a molecular modeling-AI hybrid computational approach to find potential inhibitors of the SARS-CoV-2 main protease (Mpro, 3CLpro). Naïve Bayes algorithm, an ML-based model is utilized to predict COVID-19 drugs with more than 70% accuracy. AI is being used to evaluate over 530,000 drug combinations against the SARS-CoV-2 live virus collected from a patient sample. Identif.AI identified the combination of remdesivir, ritonavir, and lopinavir as a potentially effective treatment against SARS-CoV-2 infection. Further experimental validation indicates that this drug combination exhibits a 6.5-fold enhanced efficacy over remdesivir alone. Screening drug databases and compounds have also been done using AI for drug repositioning/discovery depicted in Figure 3.6. Hydroxychloroquine and azithromycin12 were relatively ineffective against live SARS-CoV-2. Thus, Project IDentif.AI greatly cuts the number of in vitro assays required to evaluate the drug tolerability and efficacy and can be applied along with the in vitro investigations of drug validation.
A network machine learning method to target the SARS-CoV-2 host gene-gene interactome by identifying the potentially bioactive molecules in foods based on their anti-COVID-19 ability was also employed.
3.5 AI/ML in vaccine development
ML-based Reverse Vaccinology” (RV) based approaches13 viz “. Vaxign-ML and VaxiJen have been used to design a vaccine. SVM, RF, deep CNN (DCNN), and RFE, are ML-based approaches to identify the antigens from a given protein sequence for a vaccine. Vaxign and Vaxign have been used to line up NSPs as potential vaccine solutions for SARS-CoV-2. NSP3 was identified as most promising. An epitope map was created for different HLA alleles using the NEC Immune Profiler suite of tools. Tools, MARIA and NetMHCPan4, of the neural network were used to identify T-cell epitopes for the SARS-CoV-2 spike receptor-binding domain (RBD) to develop a vaccine. Simulated sequences of S protein were identified as reasonable targets using DL RNN for vaccine design. Immunoinformatic approaches were also explored to devean lop anti-peptide vaccine of S, E, and M proteins. ML-based Ellipro antibody epitope predictive method was used to predict and identify in S-protein B-specific epitopes. A combination of 19 epitope-HLA tools, using IEDB ie the Immune Epitope Database, ANN (PyTorch), and algorithm position-specific weight matrices (PSSM) was used to validate 174 epitopes of SARS-CoV-2 that could bind strongly to 11 HLA alleles.
3.6 ML for determining SARS COV 2 protein Structure
SARS CoV-2 has four structural proteins and 16 non-structural proteins. Protein Structure14 can be predicted using ML computational models. There are two modeling-based approaches to predict unknown protein structures. The template-based modeling, predicts protein structure using similar protein. SARS-CoV-2 proteins can be predicted by relating to similar proteins in some similar organisms whose proteins are known. The prediction of protein structures of proteins where no template structures are there can be determined using ML methods. DeepMind a company in the UK has developed AlphaFold to predict protein structures related to COVID-19. Its foundation is a deep neural network, ResNet architecture. It is an unbiased model protein predictor. When predicting proteins for COVID-19 it ignores similar structures, as very few related protein structures are available. AlphaFold's generated structure of SARS-CoV-2 spike protein was quite similar to results from other researchers. Alpha Fold is now being used to predict other protein structures of SARS-CoV-2 including M protein, and 3a, NSP4, NSP2, NSP6, and papain-like protease. These protein structures could then be used to discover or repurpose drugs to contain COVID-19. C–I-TASSER an extended version of I-TASSER based on deep convolutional neural network-based contact maps also known as “Zhang-Server”, is the top-most computerized server for protein structure prediction.
3.7 SARS COV 2 Protein Structure using music
ML15-based deep neural network models have been utilized to create music to represent the SARS-CoV-2 spike protein structure. ML has been used with sound (Sonification) to create SARS COV 2 structural proteins. Sonification approach “materiomusic” translates protein structures into audible signals. The actual vibrations and structures of molecules have been used to create music. Relating the structure of proteins to music, the protein hierarchical structure represents music in which the (primary sequence) of amino acids are the notes and the coil of the helix (i.e. the secondary structure) or the flatness defines the rhythm and pitch Figure 3.8 A. The overall vibrational motions of the molecules were incorporated into an audio signal using the Anisotropic Network Model. The primary signal from the structure is input into the Max device. The sounds from the proteins are generated using Ableton Live Digital Audio Workstation, which is the basis for the secondary signal. The signals from the structure and vibrations of a protein are then overlaid and played together, resulting in a multi-dimensional image of the protein's structure. Sonification of the SARS-CoV-2 S protein within a twelve-tone equal temperament tuning, produced a total of 3,647,770 notes as depicted in Figure 3.8 B. This produced a2 hours of classical music which was uploaded on the music sharing website- SoundCloud for the public to hear. The translation of proteins into a musical score involves a deep neural network. Neural Network platform (CNN/RNN) converting protein structure to music.
Using16 music and ML-enabled nanomechanical vibrational spectrum of five different protein structures of SARS COV 2, provides an insight into how genetic mutations and binding of the S protein to the human ACE2 cell receptor directly influence the audio of the music so created. The musical representations of proteins are being used to discover drugs and new therapies, discover de novo antibodies, identify druggable sites in SARS COV 2 structure, detect mutations, and material design by manipulating sound. These neural network-based musical interventions can convert protein structure to music rapidly and a database of over 10,000 protein songs is available. An android based free app called the Amino Acid Synthesizer is available to create your own protein “compositions” from the sounds of amino acids.
4. Drug discovery - Challenges
The huge volume of tests required to establish facts at the preclinical stage whicires testing of large volume of compounds. Random controlled trials (RCTs) are also needed as a precursor to drug discovery. High rates of drug attrition, associated high costs, and time delays. Reducing side-effects in clinical trials. Complex molecular-drug target evaluation.- Drug targets are in a way linked to the complex system of proteins and the molecular machinery of the cells with which they are associated There would thus be a drug-target interaction and each such drug–target interaction should be evaluated in an integrative context Figure 4.1. AI as a tool with biological knowledge from the biological systems of tissues, organelles, human interactome, etc can accelerate drug repurposing as seen on the left side of the figure. The center represents the AI component viz the computer programs and algorithms The neural network is represented by Red and Black circles. The red circle denotes neurons carrying important information from the biological systems. People represented by Green and Blue indicate subgroups that show different responses to treatment.
The downward purple arrows indicate the interaction of the AI algorithms taking information from the biological systems on the left and from the drug development pipelines o the right to build more robust models for drug repurposing.
4.1 Challenges in Drug repurposing for COVID 19
The repurposed drug would be tried on animals. But the cellular or animal assays may not reflect or replicate the host environment of the virus infection in human cells. Also, the drug targeted for repurposing may be optimized for a particular disease, tissue, dose, or target. Many drugs didn’t satisfy clinical and biological requirements because of the lack of clinical endpoints, expedient design, and lack of statistics due to the small number of patients enrolled. For example, hydroxychloroquine shows anti-SARS-CoV-2 ability in, in vitro assays but HQ was not effective in a preclinical and clinical trial. High sensitivity tools and analyses are required to detect differences between drugs and placebos, for trials with mildly affected COVID-19 patients. Different genetic backgrounds of patients affect outcomes of clinical results. Challenges in Biological interpretation -Biological systems are complex and hierarchical as depicted in figure 4.2, composed of sequences, protein complexes, cells, tissues, organs, and organisms. Drug discovery is a complicated process involving multilevel interactions between chemical compounds and biological systems. Challenges in data and model harmonization – Non-availability of a unified database, a high-quality data model. Challenges in data Sharing and Security - health-record data of COVID 19 patients, is a private data, thus concerns about security and privacy. DNA sequencing has an increased risk of a leak, as it could reveal the identity of patients. Personalized drug repurposing -SARS-CoV-2 infection has shown considerably large variabilities amongst individuals.
5. Repurposed Drug for COVID 19
The overview of19 of AI drug repurposing for COVID 19, where de-novo drug discovery is almost impossible
5.1 Repurposed drug Remdesivir
Remdesivir20 was used to treat EBOLA and to inhibit, stop the replication of coronaviruses prior. Therefore, remdesivir was the first potential drug for COVID 19 repurposing. Among other drugs evaluated in the non-human Vero E6 cells: ribavirin, penciclovir, nitazoxanide, nafamostat, chloroquine, and favipiravir. Various AI models were used to check out remedesivir as a suitable candidate for SARS COV 2. A simulated molecular docking model established remdesivir had a high binding affinity to SARS-CoV-2 RdRp. Several factors have led to the medical interest in remdesivir for the treatment of SARS-CoV-2. Remdesivir has shown in vitro activity against SARS-CoV-2 and has a safe dosage profile.
5.2 Clinical efficacy
It was determined that for clinical efficacy remdesivir 200 mg intravenously on day 1, followed by 100 mg for up to 9 more days. Remdesivir was initiated on hospital day 7 due to increased oxygen requirements and ongoing pyrexiaand in some cases even on day 13 of illness. The patient had a full recovery. Remdesivir is established to retain its therapeutic effect even if administered late in a disease course. The composition and action of Remdesivir are shown in Figure 5.1.
6. Conclusion
With the increasing availability of AI new techniques and big data (biological, clinical, and open data and scientific publications and data repositries), AI techniques21 are just ripe to be leveraged for drug discovery for treatment of new and emerging diseases. Therefore, for drug discovery and therapeutics, such large sets of biomedical data is in high demand. AI-based technologies have proven to be useful in improving drug efficacy by repurposing and aiding in the decision making of therapeutic for the treatment of complex human diseases, such as COVID-19. However, challenges as stated previously remain in developing these AI tools, due to not sharing of data by pharmaceutical companies, quality of data, its heterogeneity, security, and interpretability of the models.
References
Mohanty S, Harun Ai Rashid M, Mridul M, Mohanty C, Swayamsiddha S. Application of Artificial Intelligence in COVID-19 drug repurposing. Diabetes Metab Syndr. 2020 Sep-Oct;14(5):1027-1031. doi: 10.1016/j.dsx.2020.06.068. Epub 2020 Jul 3. PMID: 32634717; PMCID: PMC7332938.
Zhou Y, Wang F, Tang J, Nussinov R, Cheng F. Artificial intelligence in COVID-19 drug repurposing. Lancet Digit Health. 2020 Dec;2(12):e667-e676. doi: 10.1016/S2589-7500(20)30192-8. Epub 2020 Sep 18. PMID: 32984792; PMCID: PMC7500917.
Khan M, Mehran MT, Haq ZU, Ullah Z, Naqvi SR, Ihsan M, Abbass H. Applications of artificial intelligence in COVID-19 pandemic: A comprehensive review. Expert Syst Appl. 2021 Dec 15;185:115695. doi: 10.1016/j.eswa.2021.115695. Epub 2021 Aug 4. PMID: 34400854; PMCID: PMC8359727
Rasheed J, Jamil A, Hameed AA, Al-Turjman F, Rasheed A. COVID-19 in the Age of Artificial Intelligence: A Comprehensive Review. Interdiscip Sci. 2021 Jun;13(2):153-175. doi: 10.1007/s12539-021-00431-w. Epub 2021 Apr 22. PMID: 33886097; PMCID: PMC8060789.
Ke YY, Peng TT, Yeh TK, Huang WZ, Chang SE, Wu SH, Hung HC, Hsu TA, Lee SJ, Song JS, Lin WH, Chiang TJ, Lin JH, Sytwu HK, Chen CT. Artificial intelligence approach fighting COVID-19 with repurposing drugs. Biomed J. 2020 Aug;43(4):355-362. doi: 10.1016/j.bj.2020.05.001. Epub 2020 May 15
Sanghai N, Shafiq K, Tranmer GK. Drug Discovery by Drug Repurposing: Combating COVID-19 in the 21st Century. Mini Rev Med Chem. 2021;21(1):3-9. doi: 10.2174/1389557520999200824103803. PMID: 32838716
Yousefi H, Mashouri L, Okpechi SC, Alahari N, Alahari SK. Repurposing existing drugs for the treatment of COVID-19/SARS-CoV-2 infection: A review describing drug mechanisms of action. Biochem Pharmacol. 2021 Jan;183:114296. doi: 10.1016/j.bcp.2020.114296. Epub 2020 Oct 22. PMID: 33191206; PMCID: PMC7581400
National Centre for Biotechnology Information, A review describing drug mechanisms of action Available from: https://pubmed.ncbi.nlm.nih.gov/33191206/#&gid=article-figures&pid=captionless- figure-uid-0
Santus E, Marino N, Cirillo D, Chersoni E, Montagud A, Santuccione Chadha A, Valencia A, Hughes K, Lindvall C Artificial Intelligence–Aided Precision Medicine for COVID-19: Strategic Areas of Research and Development J Med Internet Res 2021;23(3):e22453
Tanoli Z, Vähä-Koskela M, Aittokallio T. Artificial intelligence, machine learning, and drug repurposing in cancer. Expert Opin Drug Discov. 2021 Sep;16(9):977-989. doi: 10.1080/17460441.2021.1883585. Epub 2021 Feb 12. PMID: 33543671.
Serena Dotolo, Anna Marabotti, Angelo Facchiano, Roberto Tagliaferri, A review on drug repurposing applicable to COVID-19, Briefings in Bioinformatics, Volume 22, Issue 2, March 2021, Pages 726–741
James Nurton, WIPO Magazine, 2020 Drug repurposing and the COVID-19 pandemic Available from: https://www.wipo.int/wipo_magazine/en/2020/02/article_0004.html
Morselli Gysi D, do Valle Í, Zitnik M, Ameli A, Gan X, Varol O, Ghiassian SD, Patten JJ, Davey RA, Loscalzo J, Barabási AL. Network medicine framework for identifying drug-repurposing opportunities for COVID-19. Proc Natl Acad Sci U S A. 2021 May 11;118(19):e2025581118. doi: 10.1073/pnas.2025581118. PMID: 33906951; PMCID: PMC8126852.
Levin JM, Oprea TI, Davidovich S, Clozel T, Overington JP, Vanhaelen Q, Cantor CR, Bischof E, Zhavoronkov A. Artificial intelligence, drug repurposing, and peer review. Nat Biotechnol. 2020 Oct;38(10):1127-1131. doi: 10.1038/s41587-020-0686-x. PMID: 32929264.
Keshavarzi Arshadi A, Webb J, Salem M, Cruz E, Calad-Thomson S, Ghadirian N, Collins J, Diez-Cecilia E, Kelly B, Goodarzi H, Yuan JS. Artificial Intelligence for COVID-19 Drug Discovery and Vaccine Development. Front Artif Intell. 2020 Aug 18;3:65. doi: 10.3389/frai.2020.00065. PMID: 33733182; PMCID: PMC7861281.
Zhou Y, Wang F, Tang J, Nussinov R, Cheng F. Artificial intelligence in COVID-19 drug repurposing. Lancet Digit Health. 2020 Dec;2(12):e667-e676. doi: 10.1016/S2589-7500(20)30192-8. Epub 2020 Sep 18. PMID: 32984792; PMCID: PMC7500917.
Kaushik, A.C. and Raj, U., 2020. AI-driven drug discovery: A boon against COVID-19?. AI Open, 1, pp.1-4.
18. Daniel Cohen Analytics India Mag, 2021, How Existing Drugs Could Be Repurposed Using AI To Fight COVID-19 Available from: https://analyticsindiamag.com/how-existing-drugs-could-be-repurposed-using-ai-to-fight-covid-19/
19. Vineeth Venugopal, Towards Data Science, 2020, Every protein is a song
A radical new way to design drugs and antibodies using music. Available from: https://towardsdatascience.com/everyprotein-is-a-song-6d30ee9addd4
20. Built a first integrated pipeline & machine learning based resource for analysis of mobile genetic elements and mutations in cancers and other diseases
Rawal, K. and Ramaswamy, R., "Genome-wide analysis of mobile genetic elements insertion sites. Nucl. Acids Res., vol. 39, no. 16, pp. 6864-6878, Sep. 2011. Impact Factor 11.3.
21. Mandal, P., Rawal, K., Ramaswamy, R., Bhattacharya, A. and Bhattacharya, S. "Identification of Insertion hot spots for non-LTR retrotransposons: Computational and Biochemical application to Entamoeba histolytica." Nucl. Acids Res., vol. 34, no. 20, pp. 5752-5763, 2006. (Lead author and equal contribution). Impact Factor 11.3.
22. Rawal, K., Dorji, S. Kumar, A., Ganguly, A. Grewal, A.S. “Identification and characterization of MGEs and their insertion sites in the gorilla genome”. Mobile Genetic Elements, vol.3, no.4, pp. e25675, 2012.
23. Rawal, K., Priya, A., Malik, A., Bahl, R., Ramaswamy, R., “Distribution of MGEs and their insertion sites in the Macaca mulatta genome”. Mobile Genetic Elements, vol.2, no.3, pp. 133-141, 2012.
24. Bakre, A.A., Rawal, K., Ramaswamy, R., Bhattacharya, A. and Bhattacharya, S., “The LINEs and SINEs of Entamoeba histolytica: Comparative analysis and genomic distribution.” Experimental Parasitology, vol. 110, no. 3, pp. 207-213, 2005.
25. Jagannadham, J., Jaiswal, H.K., Agrawal, S., Rawal, K., Comprehensive map of molecules implicated in obesity", PLoS ONE, vol. 11, no. 2: e0146759. doi:10.1371/journal.pone.0146759, 2016.
26. Jagannadham, J., Jaiswal, H.K., Rawal, K., Deciphering relationships in disease networks using computational approaches: Fatty Liver, PCOD, Osteoarthritis, cholelithiasis & hyperlipidemia", International Journal of PharmTech Research, vol. 8, no. 1, pp. 127-134, 2015.
27. Jagannadham, J., Jaiswal, H.K., Agarwal, S., Rawal, K., Biomedical Text Mining of Obesity, Diabetes and hypertension genes. International Journal of Pharmaceutical Sciences Review and Research. vol 33(2), 182-186, 2015.
28. Agrawal, S., Rawal, K., Sahu, A., Mahajan, S., Garg, P. and Bahl, R., "To find gene distributions in PubMed abstracts using Perl software", Journal of Pharmacy Research 2013.
29. Jaiswal, H.K., Rawal, K., Jaganadham, J., Agrawal, S., “Evaluation of inhibition activity of Tetrahydrolipstatin analogues on Diacylglycerol lipase alpha using In-silico techniques”. Journal of Pharmacy Research, vol.5, no.6, pp. 3473-3477, 2012.
30. Gupta, R., Soni, Patnaik, Sood, I., Singh, R., Rawal, K., Rani, V. High AU content: A Signature of Upregulated miRNA in Cardiac Diseases, Bioinformation, vol. 3, pp.132-135, 2010.
31. Rustagi Y, Jaiswal HK, Rawal, K., Kundu GC, Rani V (2015). Comparative Characterization of Cardiac Development Specific microRNAs: Fetal Regulators for Future. PLoS ONE.10(10): e0139359. https://doi.org/10.1371/journal.pone.0139359
32. Abbasi, B. A., Saraf, D., Sharma, T., Sinha, R., Singh, S., Gupta, P. Rawal, K. (2020, April 8) Identification of vaccine targets & design of vaccine against SARS-CoV-2 coronavirus using computational and deep learning-based approaches. https://doi.org/10.31219/osf.io/f8zyw.
33. Kamal Rawal, v, Abbasi, B. A., et al (2020). Design of a multi-epitope Chagas disease vaccine by computational analysis of the Trypanosoma cruzi CL Brenner proteome. (Communicated)
34. Rawal k, Sinha R et al (2020) Vaxi – DL: A web-based Deep Learning (DL) server to identify Potential Vaccine Candidates, (Bioinformatics- communicated)
COVID-19
35. Jethani B., Rawal, K. et al (2020). Clinical Characteristics and Remedy Profile of Patients with COVID-19: Retrospective Cohort Study, Accepted (In Press)
36. Rawal, K., Khurana T, Sharma H, . 2019. An extensive survey of molecular docking tools and their applications using text mining and deep curation strategies. PeerJ Preprints7:e27538v1 https://doi.org/10.7287/peerj.preprints.27538v1
37. Rawal, K., Sinha R et al (2020) To Study the Effect of Unconventional Treatment Protocol on COVID-19 patients in Delhi using AI-based techniques (Lancet- communicated)
38. Development of a Web Based Weight Loss Programme. K. Rawal, P. Gaur and K. Kashive, LAP LAMBERT Academic Publishing GmbH & Co. KG, Saarbrücken Germany, October, 2014
39. Developing Networks in Obesity using Text Mining. K. Rawal, S. Agarwal and J. Jagannadham, LAP LAMBERT Academic Publishing GmbH & Co. KG, Saarbrücken Germany, September, 2014.
40. Dev, B.B., Malik A., Rawal, K., “Detecting motifs and patterns at mobile genetic element insertion site”. Bioinformation, vol. 8, pp.777-786, 2012