Skip to main content

AI in drug discovery and development : A brief commentary

Apollo Pharmacy hiring

GPAT courses

Apollo Pharmacy hiring

pharma courses

About Author

This article is discussing about application of AI in drug discovery and development. Implementation of AI can give better results than traditional method of drug development. Discovery of a new drug is a complex and expensive process. It costs about 2.6 USD and take an average of 12 years. Application of AI technology in medical field is very important in this modern era of 4th industrial revolution where innovations and development in medical field are growing rapidly. Use of AI technology will improve efficiency of drug discovery and drug development processes. It will reduce cost, time and efforts.  It will specially help in field of “GENOME DRUG DISCOVERY” and “GENOME MEDICINE”. In drug discovery and development processes data collection and processing is very crucial part. AI can be used for development of database that will accumulate information related to disease and also relate to the cause of disease. AI in drug development has emerged as better option for traditional drug development methods.

Artificial intelligence (AI) is recently developed in a trending and sizzling topic in the world of medical care industry. AI is science of making machines that can perform tasks associated with human intelligence. AI can help so much for betterment of humanity. In present time medical care industry especially pharmaceutical industry is facing so much difficulties. Traditional drug discovery is time consuming, costly and cumbersome process. It can cost approximately 2.6 USD and can take an average of 12 years. We can reduce cost, time and difficulty by associating medical care industry (especially pharmaceutical industry) with artificial intelligence. Biopharmaceutical companies are making efforts to collaborate with AI based companies. AI based companies are also now focusing on drug discovery and development processes.

Recent advances in high-performance computing, the availability of large annotated data sets and new frameworks for implementing deep neural networks (DNNs) resulted in an unprecedented acceleration of the ?eld. DNNs have surpassed human accuracy in image, voice and text recognition, autonomous driving, and many other tasks. AI is need of hour for Need of AI

The productivity of the pharmaceutical industry is continuously decreasing. It is a matter of concern. Failure rates in clinical trials are more than 90%. Recent advances in arti?cial intelligence (including its sub?elds machine learning(ML) and deep learning(DL) may help to reduce the chances of failure of clinical trials and can boost and improve pharmaceutical development and improvement in pharmaceutical R&D.

Research in the 1960s and 1970s developed first problem solving program known as Dendral.

AI is used for mimic human intelligence for completion of tasks e?ciently, accurately and rapidly. AI in its various forms is today successfully applied across various ?elds for challenging tasks, from robotics, speech translation, image analysis and logistics to its use in designing newer molecules

AI also contains a sub?eld called machine learning (ML). which uses statistical methods with the ability to learn with or without being explicitly programmed for these tasks.
Machine learning takes arti?cial intelligence a step further. Algorithms are programmed in such a way that machines can learn and improve any data without the need for human data input and reprogramming.

ML focuses on the development of computer programs that can access data and use it to learn themselves.  ML algorithms is categorized into supervised algorithm, unsupervised algorithm, semi-supervised algorithm and reinforcement learning algorithm.

Supervised machine learning is used for prediction of future events using data learned in past. Unsupervised algorithms are used when data used to train is neither labeled nor classified. It can draw inference from a database to describe hidden structure from unclassified data.

Semi-supervised ML algorithms contain a small amount of labeled and large amount of unlabeled data. Reinforcement ML algorithm is a learning method. It interacts with environment and finds error or rewards.

A further sub?eld of ML called deep learning (DL) uses arti?cial neural networks(ANNs) that learn from the vast amount of experimental data. It is also a sub?eld of AI that processes data and creates patterns for decision making purposes. It also known as deep neural network or deep neural learning.

Deep learning is the next generation of machine learning that introduces multiple layers of learning from massive data. Deep learning decisions and data classi?cations are re?ned at each layer to produce accurate insights.

DL is emerging as a powerful tool, it is collecting massive data and getting improved day by day.  Continuing strength of DL can minimize failures in clinical trials and a faster, cheaper and effective drug development processes.

Process of Drug Development

For development of any drug require several steps. Finding molecular target to treat a disease is very first step of drug development. Target identification is done for finding a gene or protein that plays important role in that particular disease.
Hit and lead are two important terms in this process that leads drug development further. First compound that shows activity against given biological target is called hit. Process of Identifying compounds derived from hit is called hit expansion. Identification of lead is next step in drug development. Lead is also a chemical compound that is supposed to having ability to treat the disease.

After finding lead, researchers start process of lead optimization. It is done to improve modification in its chemical structure of identified chemical so that researchers can find more potent molecule that can give maximum therapeutic effect and minimum toxic effect. Experimental testing using animal efficacy models and ADMET are used for lead optimization process. When researchers find a lead compound as drug candidate, drug development starts.

After finding drug candidate drug development starts with pre-clinical research to determine efficacy and safety.
In pre- clinical research scientists determine ADME, mechanism of action, best dose, route of administration, effects on gender, interaction with other drugs, side effects, effects on gender etc.
Pre-clinical trials are done on non-human subjects for determining above parameters.

Once pre-clinical research is done researchers move to clinical drug development. It includes clinical trials and studies on human participants. Clinical trials are used to generate data about efficacy and safety. Clinical are done in three phases.
Phase I trial includes an experimental treatment of a small group of healthy people to know safety and efficacy of drug.

Phase II trial uses more people (100 to 300). It also includes study of safety and efficacy of drug but in this researchers emphasize the efficacy of drug.
Phase III trial it collects more information about efficacy and safety. it is used to study the effect of drug on different populations, in different combinations of other drugs, different doses. number of subjects ranges from 1000 to 3000.

after completion of phase III trials results available for efficacy and safety gives satisfactory results according to FDA standards, FDA approves it.

post marketing surveillance is a practice of monitoring safety of drugs after approval, manufacturing and marketing. Reports of post -marketing surveillance are submitted to FDA adverse reporting system (FEARS) database.

Development of new drug is a difficult task. This is because of a wide range of chemical compounds present in chemical space .it is estimated that approximately order of 1060 molecules are present in universe.

AI can use for easing difficult tasks e.g. drug development, drug designing, selection of population for clinical trials and drug repurposing etc.

AI reduces errors and bias caused by human in classical methods of drug development. Deep learning has given satisfactory results in finding therapeutic and toxicological effects of drug molecule.

Molecular target identification is primary and most important step for developing a new drug molecule for treatment of a disease. AI have so much datasets related to various fields and disciplines of medical sciences like genomics information, biochemical attributes etc. Target identification and validation can be done using AI and natural language processing (NLP). AI and NLP can scan a vast data of medical literature related to genetics, proteomics, genomics etc. to identify new targets. 

Use of AI for finding lead is easier than classical method of lead identification. AI in association with chemical space scientists can find possible drug molecule for treatment of disease. In advance level machine learning also helps in generation of virtual molecules that can help in treatment of disease and can find therapeutic and toxic effects virtually. In classical methods these all reactions are done in vivo and in vitro and it costs more and also time consuming. Use of AI in association with chemical space can reduce cost and time in finding hits and leads.

Molecules that obey the Lipinski’s rule of five (RO5) also known as Pfizer’s rule of five. Retrosynthesis is a very advance method for drug designing. After finding a molecule and knowing about its therapeutic and toxicological profile, next researchers need to find an optimal chemical synthetic pathway. This a very challenging task to find a pathway for synthesis of that particular compound. retrosynthesis analysis searches for backward reactions until it finds the simper and available precursor molecules. It is very hard for a human brain to process the vast number of organic chemical reactions. In modern times there are many computer organic compound synthesis systems are available for helping organic chemists. Monte Carlo tree search (MCTS) is also a very useful technique in this order. Monte Carlo tree search performs random search step without branching until it finds a simple pathway of synthesis for target molecule. This platform is much more efficient than traditional platforms. this platform is able to find an optimal route for target molecule synthesis without any unnecessary step in a very short time. Recursive neural networks (RNNs) have also been successfully used for de novo design. RNNs take sequential data as input. Transfer learning was also used in for generating novel bioactive chemical structures.

Researchers are using AI to improve clinical trials. Patient selection for a clinical trial is a crucial process. AI has a lot of medical data including electronic health records, wearable devices and many advanced machine learning and deep learning algorithms. It can help researchers a lot in various processes of clinical trials. A mobile application based on AI named AiCure was developed for II phase trial of schizophrenia. It gave a better response as compared to traditional method. AI can help in AI can save billion dollars and a lot of time. 

Pharmacovigilance is a practice of monitoring effects of medical drugs after approval. AI is a very profitable tool for pharmacovigilance. AI improves quality and accuracy of information in pharmacovigilance(PV) .AI can also manage several types of data in different formats. AI can be used for reduction in time and complexity of case processing. AI helps in identification of several Adverse drug reactions (ADRs). There are several data base tools e.g. VigiBase, VigiAccess, VigiFlow, VIgiMatch etc. also can help in PV.

Pharmaceutical industry is facing challenge in drug development programs. These challenges are increase cost of drug development and less chances of success in finding new drug molecules. AI can help pharmaceutical industry in increasing efficiency and reducing cost of drug development process.

AI tools can be used in multiple aspects of drug discovery cycle. Ai can used for finding therapeutic and toxicity effect profile of drugs, for prediction of, structure, bioactivity and mode of action of drug, selection of population for clinical trials.  AI can also help in various other fields of healthcare e.g. radiology, disease, diagnosis, gene therapy, finding drug interactions and biopharmaceutics etc.

Many new startups are growing in rapidly. Pharmaceutical companies are collaborating with AI based companies for development of newer drugs.

AI will grow further in future and it will unleash it full potential and will help pharmaceutical industry.

1. Kit-Kay Mak, Mallikarjuna rao pichika (2019) Artificial intelligence in drug development: present status and future prospect, drug discovery today, vol. 24, issue 3, pp. 773-780
2. Prashnsa Agrwal (2018) Artificial intelligence in drug discovery and development, journal for pharmacovigilance
3. Euma Ishii et al. The advent of medical artificial intelligence: lesson for Japanese approach, journal of intensive care (2020 )8:35
4. H. Chan et al. The rise of deep learning in drug discovery  Drug discovery today,23(2018) pp. 536-540.
5. P. Hamet, J. Tremblay Artificial intelligence in medicine metabolism,69(2017) pp. 536-540
6. R.C. Mohs, N.H.Greig Drug discovery and development: role of basic biological research, Alzheimer Dement. ,3, (2017) pp. 651-657
7.T. Katsila et al. computational approach in target identification and drug discovery, comput. Struct. Biotechnol. J., 14(2016) pp. 177-184
8. Segler MHS, kogej T, Tyrchan C. ,Waller M P, Generating focused molecule libraries for drug discovery with recurrent neural networks , ACS cent. Sci. 4(1),120-131(2018)
9. G’omez-Bombareli R, wei JN, Duvenand D et al.  DOGS; Reaction driven de novo design for bioactive compounds ACS cent. sci. ,4(2) ,268-276(2018)
10. Woo sung son, Drug discovery enhanced by artificial Intelligence, Biomedical journal of scientific and technical research ,12(1), 2018. BJSTR MS.ID.002189
11. H.C. Stephen Chan et al; Advancing Drug Discovery via artificial Intelligence, Trends in pharmacological sciences, vol. 40 , No. 8 , (2018)
12. Alex zhavoronkov, Artificial Intelligence for drug discovery, biomarker development and generation of novel chemistry, Molecular Pharmaceutics 2018, 15,10,4311-4313.
13. Nic Fleming, Computer-calculated compounds, Macmilan publishers ltd., part of springer nature,vol. 557, S57 (2018).
14. Arun Das, Paul rad, Opportunities and challenges in explainable artificial intelligence (XAI): A survey.
15. Matthew a sellwood, Artificial Intelligence in drug Discovery; Future med. Chem. (2018)10(17),2025-2028.
16. Clemence Reda et al, Machine learning applications in drug development, Computational and biotechnological journal (2020)241-252.
17. Kotani Murali, Sukhmeet Kaur, Ajay Prakash, Vikas Medhi ; Artificial Intelligence in pharmacovigilance: Practical utility, Indian J Pharmacol. Nov-Dec; 51(6): 373-376.





$(document).ready(function() { $(document).on('click', "[data-toggle='collapse']", function(event) { $(this).toggleClass('collapsed'); var targetval = $(this).attr('data-target'); $(targetval).toggleClass('in'); }); });