Abstract
The articles in this special collection address topics on the use of artificial intelligence (AI) techniques and technologies as applied to drug discovery, automated gene editing with a single-cell platform using computer vision, and a specific application of machine learning (ML) breaching the biorelevance gap in antioxidant assays.
AI has been finding more applications in drug discovery, development, and the manufacturing value chain in recent years. This is due to significant advances made in the past 20 years in computational power, cloud storage, and computing; increased accessibility to ML and deep learning platforms; and connectivity to big data, to name a few. While we still need to understand what is practically possible to achieve using AI versus aspirations, there are areas in medicine in which AI techniques are successfully applied, such as computer vision and ML using big data.1,2 In our own research and industry experiences, we were able to implement ML and deep learning to drug manufacturing process monitoring,3–5 as well as automating microscopic cell image detection in process development. Including our applications, a broad array of AI use in drug discovery, development, and manufacturing was shared at a recent AI conference at MIT. 6 Regulatory authorities are also progressing toward more adoption of AI technologies in drug development, medical devices, and manufacturing applications.7,8
In the first article in this special collection, Rashid 9 provides a review of AI in drug development summarizing techniques and approaches used as the current state of the art. The author highlights the potential of AI and argues that the AI-driven automation advances will continue in the areas of novel drug target identification and design, biomarker identification, and effective patient stratification.
The second article, by Idowu and Fatokun, 10 is a specific application of ML methods in the important area of biorelevant antioxidative assays. They developed a predictive tool using various ML methods for assessing the biorelevant antioxidant capacity of polyphenols to identify and design antioxidant molecules. The authors offer a balanced view of the advantages and drawbacks of using ML, namely, the challenge of having a limited amount of datasets in some of the experimental areas.
One of the most mature areas of AI is image processing and computer vision using deep learning. A great demonstration of computer vision application using deep learning for automated gene editing using a single-cell electroporation platform is provided by Patino et al. in the third article. 11 This study is a good example of automation of high-throughput assays augmented with powerful deep learning methods to achieve the efficiency and speed required in cell manipulation and execution of tasks.
In conclusion, this selection of articles provides a good summary of AI applications as a representative set of the current state of the art of AI in select areas of bio/pharma, and it shows the realm of the possible in coming years in which we anticipate seeing more applications of AI in the discovery, development, and manufacturing of new medicines.