3 Reasons Artificial Intelligence is the Future of Drug Discovery

Written by: Noah Parekh

The role of artificial intelligence (AI) in the biopharma industry has been growing geometrically for the last several years. New applications appear on a near-monthly basis, from automated manufacturing to market analysis, clinical trial design, QA and QC, and compliance monitoring. As a result, drug discovery is among the most exciting – and fastest-growing – domains for its use. There are at least three reasons AI is likely to be central to identifying and evaluating most or all new candidate drugs within the next decade. 

First, AI allows for tremendous savings in both time and cost. An early leader in this field was Insilico Medicine, which published some of the first works in generative medicine in 2016. They have recently developed a proprietary deep learning system (Generative Tensorial Reinforcement Learning, or GENTRL) and applied it to identify novel DDR1 kinase inhibitors for use in anticancer drugs. A recent paper reports that GENTRL generated six candidate molecules, which were then narrowed to two candidates using biochemical and cell-based assays. The entire process took a month and a half. 

Nor is Insilico an isolated success. In early 2021, a similar process allowed the German biotech Evotech – in collaboration with UK-based Exscientia – to move a new anticancer drug to clinical trials in 8 months, down from an anticipated 4 to 5 years using traditional methods. 

The story is much the same when we look at finances. The total pre-market cost of launching a new drug is estimated at $2.6 to 2.8 billion, with hundreds of millions more required for research and development once approval is secured. Some analyses suggest that AI might reduce that cost by as much as 70%. 

Savings in both cost and time result from efficiency, automation, and scale. The use of deep learning and other forms of AI changes the earliest steps of drug discovery from a target-driven search where human chemists test thousands of candidate molecules into an automated, data-based process where customized machine learning algorithms can run simulated tests (and other forms of simulated analyses) to identify lead-like molecules. Often, the process can be completed using existing data. Eventually, this process may evolve to a point where AI systems can suggest the structure of new therapeutic molecules from scratch. The tech doesn’t exist yet, but it is undoubtedly on the horizon. 

These same methods create a second significant benefit for AI-driven discovery processes: they can identify potential risks of candidate drugs that would otherwise be missed. We have seen this in action already, with early-generation pharmaceutical AI successfully characterizing the metabolic pathway for terbinafine, an anti-fungal medication that had caused several deaths due to liver failure. The metabolic process in question was sufficiently complex that it had eluded human investigators until deep learning was leveraged. Here, too, companies can rely on existing datasets – including the federal Tox21 program – as input for their algorithms, quickly expanding their ability to screen for toxicity and other risks. 

A final major benefit afforded by the kinds of deep learning systems in play here is that, bluntly, they can identify candidate molecules that humans will likely never have the ability to find. The drug discovery methods our industry has relied on for the last century are no longer adequate: every year sees fewer innovative drug targets, fewer new mechanisms discovered, and fewer first-in-class drugs brought to market (or even hypothesized). The amount of work required to make these advances is increasing, both in terms of time and resources and sheer computational power. AI offers robust, scalable solutions to that problem, allowing us to escape the limitations of human cognitive capacity without sacrificing human oversight or expertise at later stages of the discovery process.

Over the coming months, we will offer further insights into the current state of the AI market, examining major players and the technology that is driving the field forward. 

To discuss this further, we invite you to reach out to Noah Parekh.