2nd June 2021Feature
How is AI being used in early drug discovery?
Blog written in partnership with Algorithm379
The pharmaceutical industry is looking for ways to streamline drug development in order to reduce capital expenditure and increase success rates. One major trend in the industry in this regard is increasing use of artificial intelligence (AI) and machine learning (ML). Advances in machine learning, particularly in complex neural network algorithms, have enabled machines to learn from existing data in such a way that algorithms can often predict successful results from data not seen before. ML/AI applications currently rely heavily on the element of human touch used to program them and their success is largely dependent on how well a given question can be formulated for an algorithm.
AI/ML offers tools to sift through large volumes of biological data, discarding the noise and selecting clinically meaningful features, which not only saves time in R&D but also significantly reduces costs. As a result, there is a growing trend of leveraging the use of AI in drug discovery and development, including in target identification, lead optimization, and selection of preclinical cancer models and patient populations for clinical trials and many other milestones of the drug development process.
Focusing on therapeutics pipeline development, Artificial Intelligence and Machine Learning can typically come into play at two different stages:
- Therapeutic target identification
- Compound development
Traditionally, therapeutic target identification relied on many in vitro assays of increasing scale and complexity trying to pinpoint the exact disease mechanisms and identify a single target or a set of targets that needed to be modulated in order to reverse or alleviate the diseased state. However, there is a clear shift happening in this paradigm with bioinformatics, computational biology and/or cheminformatics increasingly orchestrating target selection based on mined screening data.
Compound development involves deciding on a therapeutic strategy (e.g. small molecule compounds, biologics or combinations of both, etc.). This discovery section has also become more reliant on computational methods in recent years, where compound libraries are designed and screened in silico in a high-throughput manner. All of this culminates in a set of candidate hits that can go into high-throughput in vitro screens to validate efficacy and fine tune molecular properties.
All of these analyses are necessary to safely and effectively progress to clinical studies.
Want to learn more about how AI/ML is being used in the early drug discovery cycle?
Download Part 1 of the AI in Drug Discovery and Development eBook here.
E-book contents include:
- Current state of the pharmaceutical industry
- Reasons for low success rate of drug discovery
- Definition of Artificial Intelligence and more details on ML and DL
- Role of AI and ML in drug discovery, including in target discovery and in silico assays