Drug Discovery In The Age of AI

Previously, we contemplated the practical opportunities of the assistance of artificial intelligence (AI) technologies in clinical trials for drug development. Such assistance extends before and after trials commence, and in this article, we will focus on the pre-clinical trial aspect.?

In particular, we will focus on the drug discovery stage which involves the identification and creation of new medications. This process has, traditionally, been a time-consuming and labour-intensive one. In the age of AI, drug discovery can be made more efficient and precise. We will take a look at how this can be the case in this article.

The drug discovery process

The drug discovery process can be considered as the first stage in drug development. This usually starts with the identification of a target protein that is associated with a disease. Once identified, “high-performance selection” is performed to find molecules that can bind to and influence the target. These molecules become the initial drug candidates.

These candidates then need to be optimised to enhance their binding to the target. Afterwards, they are characterised to understand their actions in the human body. Following these steps, a successful molecule can proceed to further drug development stages.

Traditionally, the drug discovery process has involved considerable trial-and-error research before a drug can proceed to further developmental stages. Now, AI technology can enhance the process.

AI’s role in drug discovery

The assistance of computers and mathematical models in designing new drugs stretches as far back as the 1970s . However, in the current age of AI that we live in, such assistance has taken a new dimension. Such smart algorithms can aid virtually every step in drug discovery and beyond. In the past decade alone, the amount of AI-discovered drugs has significantly increased. In 2023, 46 reached phase II and III clinical trials.?

We dedicated an article to the assistance of AI in clinical trials . But before reaching these stages, AI can also provide assistance. Here, we will consider the practical side of the drug discovery stage with some concrete examples.

Identifying targets

The analytic prowess of AI models enables them to scan considerable amounts of datasets , from genomic to clinical data. This, in turn, allows them to precisely identify targets and how potential drugs will interact with them, in a time-efficient manner.?

London-based Benevolent AI leverages this ability to analyse data from the scientific literature, as well as clinical and chemical databases. Through such an approach, they could develop a drug to treat ulcerative colitis that is undergoing clinical trials.

“Based on the information that’s provided, [the AI model] is able to almost think and propose something that was previously unknown,” Anne Phelan, chief scientific officer at London-based Benevolent AI, explained to The Pharmaceutical Journal .

Drug formulation

A drug’s formulation, or its chemical structure and composition, determines its effectiveness in practice. Algorithms can be used to predict these characteristics of a candidate and whether it will be effective and safe.

Drug discovery software developer Schr?dinger employs AI models to predict how molecules will behave and the possible outcomes, based on their respective parameters. The company’s tools have been adopted by pharma companies such as Pfizer and? AstraZeneca.

Another promising AI tool for drug formulation is Google’s AlphaFold . The model can accurately predict how potential candidates bind to proteins and influence human disease. Isomorphic Labs utilises AlphaFold and its own AI models to improve drug design for projects with its pharmaceutical partners.

Drug repurposing

As the term suggests, drug repurposing involves employing existing drugs, initially developed for other uses, in new medical applications. This approach can fast-track drug discovery cost-effectively.?

Some companies are using AI for such purposes. For example, US-based Recursion has three repurposed drugs undergoing clinical trials and has partnered with Genentech and Bayer for drug discovery.

Technological limitations and ethical challenges to be acknowledged

Despite the undeniable advantages of AI in aiding drug discovery, there are some limitations to the technology that need to be acknowledged. In particular, the value and reliability of the insights from such models rely on the data used to train them. Some conditions or medical fields such as oncology have more research input and available data; while others, such as tropical diseases, might not share similar levels of available information. Such gaps in data can be filled over time, but this will require patience and adequate resources.?

Furthermore, the ethical considerations in using such tools must not be overlooked. Companies developing AI for drug discovery must ensure the ethical use of training data and that the outputs are not biased in order to have reliable real-world use.