The Potential of AI in Drug Discovery and Development

Laiba Saleem, Aligarh Muslim University

Artificial Intelligence (AI) is not any doubt the core way forward for IT, and together with machine learning and data science it holds significant potential in majority fields, and the medicinal industry is not any stranger thereto. From medical/surgical robotics to microchips containing the anamnesis of patients, AI has come a long way and is well established to shape-change the medical field.

Limitations to Traditional drug discovery

Traditional drug discovery methods are indeed time-taking and cumbersome processes. These work effectively for druggable targets which manifest defined structures and interactions. The research of certain undruggable targets is restricted because of the complex nature of cellular interactions and limited knowledge of intricate cellular pathways. Pharmaceutical companies routinely face drug development timelines of around 15 years, sometimes costing way over $1 billion with very minute rates of success. It’s estimated that 1 in every 10 small molecule projects become candidates for clinical trials and about 1 in 10 of these compounds then successfully tolerate all clinical trials. The shortcoming to plan a molecule that selectively drugs the specified target and insufficient financial incentives that meet the scale of the addressable market results in computational aided drug discovery.

AI and Drug Discover

The research in AI applied to chemistry has largely been fuelled by the requirement to accelerate drug discovery and reduce its huge costs and also the time to promote for brand spanking new drugs. Artificial Intelligence network systems including Artificial Neural Network (ANN), Multiple Layer Perceptron (MLP), Convolutional Neural Network (CNN), Natural Language Processing (NLP) aids in discovery, designing and development of lead compounds.

• Artificial Neural Network (ANN) is a computational non-linear model supported the complex body part of the brain which is capable of performing operations like classification, prediction, decision-making, and visualization has helped in recognizing hit and lead compounds and provides a quicker validation of the drug target and optimization of the drug structure design effectively and efficiently.
• Multiple Layer Perceptron (MLP) includes three or more layers that use nonlinear activation functions such as hyperbolic and logistic functions and analysis is not linearly separable. MLP with natural language processing (NLP) aids in speech recognition and machine translation.
• Convolutional Neural Networks (CNN) as the name suggests, uses one or more complex layers which are pooled or fully connected using multilayer perceptron. These layers use complex operations as input and mapping, interpretation is carried out in subsequent convolutional layers, and desired or matching output is given out. The applications are image and speech processing and recognition.
• Natural Language Processing (NLP) is a branch of artificial intelligence that serves as an interaction between computers and humans using natural language. Its main purpose is to read, convert, interpret, and make human language sensible and valuable to computers. The area that NLP covers in novel drug discovery includes-NLP based literature curation i.e., with help of NLP the researchers can easily find the most accurate and precise data, collection of peer-reviewed data which provides evidence that helps in assessments of the quality of research finance requirements. NLP helps in identifying domain knowledge of unstructured/ unknown compounds by text-based representations of already known databases and thus by exploring the codon and bases pharmaceutical companies can easily identify noble entities or drug targets of compounds which are unknown to bind to already existing/proven drugs. NLP also helps in mining the records or medical transcripts of patients especially in clinical trials which helps in getting effective and efficient clinical trial results. NLP also allows researchers to picture how drug/drug targets act on patients, the effects or symptoms of its administration concerning existing databases.

The potential of AI in drug discovery

AI doesn’t depend upon predetermined targets for a leading compound. Therefore, subjective bias and existing knowledge isn’t an obstacle in the development process. Since AI is data-driven and complicated algorithms and machine learning extract meaningful data from large data-sets to develop state-of-art algorithms for drug discovery likewise, paving the way for AI to level the playing field in the drug development process. Because of the higher predictive power of AI in defining meaningful interactions in an exceeding drug screen, the potential for false positives is often reduced by carefully designing parameters of the assay. The potential of AI to maneuver drug screening from the bench to a virtual lab has been proven as a boon to scientists, where results of the screen may be obtained with greater speed and promising targets will be shortlisted without the necessity of in-depth experimental inputs and manual hours.

An instance of how AI is used in the identification of drug targets

The revolution of the contemporary drug discovery process to AI and large data sets discovery is due to the advancements in the field of microarray, RNA sequencing, high-throughput sequencing (HTS), and large biomedical datasets engendering. The discovery process starts with the identification of appropriate targets (proteins or genes) which are responsible for disease pathophysiology and then finding the lead compounds or targets which can interfere with the identified targets. Using gene expression, a researcher can find out the targets responsible for the pathophysiology. RNA sequencing and Microarray technologies provide gene libraries or gene repositories where a large amount of gene expression databases is stored. Some gene repositories include NCBI Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/), The Cancer Genome Atlas (TCGA) (https://www.cancer.gov/about-nci/organization/ccg/research/structural-genomics/tcga), and Array Express (https://www.ebi.ac.uk/arrayexpress/). By interpreting and analyzing the gene data sets researchers can identify genes responsible for the particular disease. Van IJzendoorn et al. 2019 interpret and find out the gene expression (biomarkers and drug targets) responsible for a rare soft tissue sarcoma, using ML and gene expression data sets.

Instances of research thrust in AI and drug discovery are listed below:

AI has made a huge difference in pharmaceutical research on thrust areas including aggregating and synthesizing information, understanding disease mechanisms, and generating novel drug candidates

1- Aggregating and synthesizing information

• Combines new RNA sequencing technologies with proprietary machine learning.

• Mine data to quickly identify the direct targets of a unique drug.

2- Understanding disease mechanism

• Analysis of genome-wide screens.

• Identifying proteins involved in regulating the cell cycle.

• Discoveries of next-generation therapies against cancer.

• Training computer vision ad machine learning models on cryo-EM (Cryo-Electron Microscopy) data.

• Provide the detailed spatial 3D structure of proteins and molecular complexes.

• CryoSPARC (Cryo-EM Single-particle Ab-Initio Reconstruction and Classification) -system software enables reconstructions of research and drug targets.

3- Generating novel drug candidates

• Structure-based deep CNN

• Predicting bio-activities of small molecules.

• Screen compound libraries for efficiency against diseases.

• Identify biological targets.

• Uncover novel disease biology hypothesis supported by world data.

Challenges with the Implementation of AI

Artificial Intelligence with machine learning uses algorithms to interpret, investigate and give out results using large data sets. So, in these cases, the algorithms should clearly and transparently layout datasets that can be used to analyze and interpret useful and meaningful information where results are in the Gray-zone of interpretation. ‘Algorithm – bias’ is another challenging drawback for AI where creators manipulate certain data sets and the software will not be able to evaluate the desired input and thus manifestation is biased and is not entirely objective. The most challenging hurdle in the implementation of AI is the finance of super-computing, libraries of large data sets, and high-throughput screening. In addition, these highly skilled computational scientists are required to check and verify the screening and predictions. Although there aren’t any drugs currently within the market developed with AI-based approaches, specific challenges continue concerning the implementation of this technology.

In silico nominated their small molecule inhibitor, in December 2020 for an investigational new drug (IND) implementing studies and expecting clinical trials by early 2022. If the trials are fruitful, then it will be the AI-based first-ever novel target and its inhibitor that got approved. AI will likely become a useful tool within the pharmaceutical industry soon.

Also read:Paleontological studies interpose Dragon Man as our sister lineage

References:

1. Jiménez-Luna, J., Grisoni, F., Weskamp, N., & Schneider, G. (2021). Artificial intelligence in drug discovery: Recent advances and future perspectives. Expert Opinion on Drug Discovery, 1–11. https://doi.org/10.1080/17460441.2021.1909567
2. Azofeifa, J. G., & Dowell, R. D. (2017). A generative model for the behavior of RNA polymerase. Bioinformatics, 33(2), 227–234. https://doi.org/10.1093/bioinformatics/btw599
3. Van Drie, J. H., & Tong, L. (2020). Cryo-EM as a powerful tool for drug discovery. Bioorganic & Medicinal Chemistry Letters, 30(22), 127524. https://doi.org/10.1016/j.bmcl.2020.127524
4. Arabi, A. A. (2021). Artificial intelligence in drug design: Algorithms, applications, challenges and ethics. Future Drug Discovery, FDD59. https://doi.org/10.4155/fdd-2020-0028

About Author:

Laiba Saleem is an undergraduate with a keen interest in the area of medicinal and computational chemistry particularly drug design and development. She is pursuing B.Sc. focused in Industrial Chemistry from Aligarh Muslim University, in Aligarh.

Social media link:

LinkedIn: https://www.linkedin.com/in/laiba-saleem-3218b91bb

Instagram: laiba___siddiqui

• The Corrosion Prediction from the Corrosion Product Performance
• Nitrogen Resilience in Waterlogged Soybean plants
• Cell Senescence in Type II Diabetes: Therapeutic Potential
• Transgene-Free Canker-Resistant Citrus sinensis with Cas12/RNP
• AI Literacy in Early Childhood Education: Challenges and Opportunities