From bench to bedside: how AI is turbocharging drug discoveries

Martin Sandhu

March 2024

The process of bringing a new drug to market is notoriously long, risky, and expensive. By some estimates, it takes over 10 years and costs upwards of £2 billion to go from initial research to an approved therapy. This sluggish pace deprives patients of life-changing treatments and threatens the sustainability of the biopharmaceutical industry, especially that of start-ups or smaller independent outfits. However, the recent emergence of artificial intelligence (AI) technologies is set to radically transform this status quo.

Across every step of the pipeline, from discovery through clinical trials, AI is making drug research and development faster, cheaper, and more effective. As this technology matures, it promises to shave years off development timelines and enable breakthrough treatments for previously untreatable diseases. The implications for patients are profound, especially those suffering from age-related conditions like cancer, Alzheimer’s and heart disease.

In this article we’ll examine the pivotal role AI could potentially play in getting vital new medicines from laboratories to patient in an efficient and cost-effective manner.

Turbocharging early-stage drug discovery

The initial steps of identifying and validating therapeutic targets have always been risky, expensive bottlenecks in development. Yet this is changing fast thanks to AI and automation. Rather than manually screening hundreds of thousands of molecular compounds through trial-and-error, algorithms can now predict the most promising candidates with stunning accuracy.

For example, companies like BenevolentAI and Exscientia are using AI to hunt for targeted treatments by scanning vast databases of research. Their platform analyses connections across scientific papers, clinical trials data, chemical structures and more to pinpoint where existing compounds may be effective. This “knowledge graph” approach is phenomenally successful at surfacing hidden patterns human researchers would likely overlook. What’s more, AI can produce convincing literature reviews of existing scientific papers and summarise them for users. This action can help speed up the research process and save time and energy for already stretched R&D teams.

Other firms like Absci are going even further by leveraging generative AI models like AlphaFold. These models can accurately predict 3D protein shapes to help design novel drug compounds with specific binding properties from scratch. This eliminates the need to manually screen libraries of molecules one-by-one in the lab. Alongside ultra-fast robotic experimentation systems, the entire early discovery process can be exponentially streamlined, which sends savings straight to the bottom line.

Driving more efficient clinical trials

AI is also making enormous inroads around clinical testing, which remains an uncertain, tedious affair. It typically takes 6-7 years to go from early trials through late-stage human studies. Failure rates also hover above 90%, which wastes precious resources. Here too, algorithms promise to drastically improve matters.

Platforms like Trials.ai and Deep6 are using troves of historical data to better predict real-world trial outcomes. This allows them to simulate trials with digital patients and refine protocols before a single person receives treatment. By modelling elements like trial enrolment criteria, dosing, or care regimens, studies can be optimised on computers first. This takes much of the guesswork out of trial design to reach endpoints faster.

Other companies like Unlearn.AI can also sniff out sources of bias or variability in studies, which frequently render them useless. Their models surface hidden factors skewing control groups or measurements to maintain experimental integrity. And tools from Saama, Biovista and other AI trailblazers help uncover leading indicators during trials to make mid-course corrections on dosages or switch up treatment arms.

Together, these techniques stand to cut development timelines by one to two years with significantly fewer patients than traditionally studies needed. The downstream effect is faster access and lowered costs for pharmaceutical firms, allowing resources to be channelled into new research programs.

Slashing the time from lab to launch

While AI has untapped potential across R&D, its benefits are clearest heading into the backstretch before regulatory approvals. This gauntlet of formulating, scaling and validating the manufacturing process causes substantial delays. However, innovators like Strateos and Snapdragon Chemistry are demonstrating how AI both streamlines and de-risks commercialisation.

By gathering data from hundreds of lab-scale production runs, algorithms can pinpoint the optimal conditions for scaling. The manufacturing process involving precise temperature, solvent and stir settings is also largely automated using robotic cloud labs. This allows chemists to remotely control reactions from design libraries instead of manually constructing countless test batches. Hence, the fastest, most reproducible formulations rise to the top.

Once synthetised qualifying therapies using AI simulation models and genomics testing further ensures batch consistency. This fulfils legal stability and purity requirements while generating air-tight data packages and documentation for regulators. Augmenting submissions with AI auditing throughout development also guarantees fidelity, satisfying compliance in one swift review.

Leveraging algorithmic insights for precision manufacturing and regulatory prep reduce timeframes that ordinarily eat up valuable years. Treatments deemed efficacious in trials can cross the finish line in record time.

The outlook for faster cures

Industry observers estimate AI could cut more than five years from a typical development cycles by mid-decade while slashing costs up to 50% or more. The broad expectation is that instead of the glacial pace of progress typified by past blockbusters, new drugs and treatment regimens will rollout at unprecedented speed. Patients in desperate need stand to be the greatest beneficiaries.

Nowhere will this be more crucial than with diseases closely correlated with aging like cancer, heart disease and neurodegeneration. As average lifespans lengthen in developed nations, finding innovative therapies for challenging age-related conditions grows more urgent by the day. AI is already uncovering novel targets and pathways for intractable illnesses, including the deadliest brain cancer glioblastoma. The same goes for accelerating clinical programs for Alzheimer’s, ALS and similar neurodegenerative scourges affecting millions globally.

Bolstered by algorithms, R&D labs racing towards disease-modifying or even preventative drugs for such illnesses are gaining invaluable momentum. While it is still early days, AI undeniably serves to amplify and accelerate science underlying tomorrow’s breakthrough medicines. For patients anxiously tracking progress on devastating diagnoses, this gives real hope that revolutionary treatments lie just over the horizon. The era of harnessing AI to speed the march of therapeutic progress from bench to bedside has clearly arrived.

If you would like to learn more about the impact AI and new technologies are having on our health as patients, download our latest white paper on the most significant trends of the year: https://www.nuom.studio/trends

Like this?

More

AI

insights

View more insights

Let’s talk

We create human-centered solutions that drive positive outcomes for users and organisations. Let’s collaborate.

See our work