The Use of AI for Enhanced Pharmaceutical Drug Discovery 

Introduction

 

Drug discovery is a time-consuming and costly endeavour; the mean cost of bringing a drug to market is estimated between $314 million to $2.8 billion (Wouters et al., 2020) and takes on average 12 years (Moingeon et al., 2022). Although the impact of AI implementation is hard to quantify, it’s use will undoubtedly increase the efficiency, accuracy and success of this process, reducing pharmaceutical companies’ costs and saving countless lives.

​

Current Drug Design Process and Limitations

​

The current drug design process involves target identification, prediction of the target’s druggability, drug screening, lead identification and subsequent pre-clinical and clinical trials (Zhuang & Ibrahim, 2021).Traditionally, identifying lead compounds relies on labour intensive techniques such as trial and error and high-throughput screening (HTS) (Blanco-González et al., 2023). HTS involves rapidly testing a library of chemical compounds that are screened using an assay to identify ‘hits’ (usual hit rates of 0-0.1%) that show promising activity for the target molecule. The appeal of AI arises from its ability to identify patterns and trends in vast HTS datasets that may not be apparent to human researchers (Blanco-González et al., 2023), enabling the prediction for the activity of novel compounds more accurately (Blanco González et al., 2023).

 

AI and its Capabilities in the Pharmaceutical Industry

 

AI involves several method domains, such as reasoning, knowledge representation, solution search and the fundamental aspect of machine learning (ML) (Paul et al., 2021). ML uses algorithms to recognise patterns within a set of data and predict future data, whilst also showing great strength in solving complex mathematical problems (Selvaraj et al., 2022). Within ML, lies a sub-field known as deep learning (DL), focussing on neural networks that are capable of learning hierarchical representations of data (Selvaraj et al., 2022). Artificial neural networks (ANNs), are composed of three layers (input layer, hidden layer and output layer); within each layer exists a single computing unit (known as neurons) that performs non-linear transformations of data (Selvaraj et al., 2022). Molecular modelling and drug design rely heavily on ANNs; their ability to resolve complex issues associated with statistical modelling in HTS shows promise in their use in medicinal chemistry (Selvaraj et al., 2022). Furthermore, ANNs perform excellently in generating numerical values to interpret non-linear relationships and predict the success of certain drugs (Selvaraj et al., 2022).

 

Case Study using Deep Learning AI (Zhuang & Ibrahim, 2021)

 

Zhuang and co. investigated the use and effectiveness of a DL model to identify and validate high efficiency drug compounds for the treatment of SARS-CoV-2 viral cells. The team used the pre-existing DL Model DenseNet to perform transfer learning, in which a pre-trained model in a specific domain is refined to be used in a different domain. They made use of 2 datasets from Recursion: RxRx1 and RxRx19a. They initially trained the model with RxRx1, a dataset based on human renal cortical cell (HRCE) treatments of 1138 classes of siRNA. Three different types of HRCE cells were used: mock cells (the control), cells with a deactivated SARS-CoV-2 virus and active SARS-CoV-2 cells. They then applied transfer learning a second time using RxRx19a, a dataset with similar characteristics to RxRx1, but involving cell treatments of a range of compounds from a library of drugs, generating 305,520 morphological images. A Softmax layer was applied each time to produce probability scores, allowing for statistical analysis. They compared and validated their results using other models as well as FDA approved drugs, arriving at similar conclusions. Although this study is not necessarily applicable to all research due to the limitation of obtaining similar datasets, the results are encouraging and highlight the use of AI in speeding up the process of drug selection and prioritisation.

 

Challenges and Limitations Associated with AI

 

Despite the potential benefits of AI use, there are several challenges that must be considered. AI-based approaches require input of large volumes of information and data for training purposes, which can often be hard to obtain due to private ownership of data by pharmaceutical companies in a competitive market (Schneider et al., 2020). This issue is partly resolved by the continued digitalisation of data and research but remains a primary barrier to the use of AI models. Furthermore, there is a concern over the quality of data used and the potential for bias (Vora et al., 2023), as well as worry over data privacy and security. To negate these problems, increased AI regulation is imperative, ensuring AI models are trained with diverse and representative data, as well as implementing strong data privacy protocols. Conclusion As increasing amounts of data become accessible for AI-training purposes, the use of these techniques has the potential to revolutionise the drug discovery process, making it faster, cheaper and reducing failure rates. Increased regulation of these models is necessary before the healthcare industry can smoothly integrate them into practise and research. Overall, better regulated AI with the combined use of human research expertise, has the power to optimise the drug discovery process.

​

 

References

​

Blanco-González, A., Cabezón, A., Seco-González, A., Conde-Torres, D., Antelo-Riveiro, P., Piñeiro, Á., & Garcia-Fandino, R. (2023). The Role of AI in Drug Discovery: Challenges, Opportunities, and Strategies. In Pharmaceuticals (Vol. 16, Issue 6). https://doi.org/10.3390/ph16060891 Moingeon, P.,

 

Kuenemann, M., & Guedj, M. (2022). Artificial intelligence-enhanced drug design and development: Toward a computational precision medicine. In Drug Discovery Today (Vol. 27, Issue 1). https://doi.org/10.1016/j.drudis.2021.09.006 Paul, D., Sanap, G., Shenoy, S., Kalyane, D., Kalia, K., &

 

Tekade, R. K. (2021). Artificial intelligence in drug discovery and development. In Drug Discovery Today (Vol. 26, Issue 1). https://doi.org/10.1016/j.drudis.2020.10.010 Schneider, P., Walters, W. P.,

 

Plowright, A. T., Sieroka, N., Listgarten, J., Goodnow, R. A., Fisher, J., Jansen, J. M., Duca, J. S., Rush, T. S., Zentgraf, M., Hill, J. E., Krutoholow, E., Kohler, M., Blaney, J., Funatsu, K., Luebkemann, C., & Schneider, G. (2020). Rethinking drug design in the artificial intelligence era. In Nature Reviews Drug Discovery (Vol. 19, Issue 5). https://doi.org/10.1038/s41573-019-0050-3 Selvaraj, C., Chandra, I., &

 

Singh, S. K. (2022). Artificial intelligence and machine learning approaches for drug design: challenges and opportunities for the pharmaceutical industries. In Molecular Diversity (Vol. 26, Issue 3). https://doi.org/10.1007/s11030- 021-10326-z Vora, L. K., Gholap, A. D., Jetha, K., Thakur, R. R. S.,

 

Solanki, H. K., & Chavda, V. P. (2023). Artificial Intelligence in Pharmaceutical Technology and Drug Delivery Design. In Pharmaceutics (Vol. 15, Issue 7). https://doi.org/10.3390/pharmaceutics15071916

 

Wouters, O. J., McKee, M., & Luyten, J. (2020). Estimated Research and Development Investment Needed to Bring a New Medicine to Market, 2009-2018. In JAMA - Journal of the American Medical Association (Vol. 323, Issue 9). https://doi.org/10.1001/jama.2020.1166 Zhuang, D., & Ibrahim, A. K. (2021). Deep learning for drug discovery: A study of identifying high efficacy drug compounds using a cascade transfer learning approach. Applied Sciences (Switzerland), 11(17). https://doi.org/10.3390/app11177772