One in every 1,000 of the drugs that begin preclinical testing make it to human testing. That’s over 90% of potential medications being ruled out at the earliest stages of trials.
Soaring drug-discovery costs and testing times only add to the gauntlet that new drugs need to run before they can get on the market and reach the patients who need them most.
However, artificial intelligence may be opening up another way to get medication on the market, dramatically cutting the time it takes to find and develop viable drugs with fewer risks. By searching vast databases of potential molecules, robots can algorithmically calculate which will be the most effective treatments for a particular disease.
Leo Barella, Global Head of Enterprise Architecture for AstraZeneca, has gone as far as to suggest that AI will become the primary tool for drug discovery by 2027.
With artificial intelligence rapidly reshaping the future for researchers and patients alike, we look into how these sophisticated AI systems work, and how they can drastically improve efficiency across big pharma.
AI Drug Discovery
BenevolentBio is one drug company that aims to develop deep-learning AI, capable of selecting potential medications. Their programs will analyse bioscience information from across patents, genomic data, and medical journals and databases, processing this information for retrieval when a relevant drug is needed.
So far, BenevolentBio have used deep learning to find two possible treatments for ALS. Using supercomputers, the software sifts through vast quantities of information to find connections between potential molecules and specific diseases, resulting in a list of potential drug candidates.
In early tests, the suggested drugs worked better than those already on the market.
Current drug development procedures involve manually screening large numbers of molecules for potential drugs, then testing each option painstakingly over the course of several years, with the hope of finding one that ticks all the boxes.
With AI in the picture, however, systems can ‘imagine’ the ideal molecule, with specific properties that will tackle the disease in question. Potential molecules can then be held up against the template of the imaginary molecule to see how well they fit the brief.
This process of discovery also makes it easier to avoid adverse effects during the trial period, as the software’s algorithms can rule out any potential drugs that will affect other parts of the body.
Leo Barella’s statement that use of AI for drug discovery will become mainstream by 2027 was made in the context of a more individual approach to drug development.
He noted that not all drugs work for everyone, and that the pharmaceutical sector needed to increase its focus on individual patients, rather than making assumptions about a particular disease.
With AI at the helm, however, individual differences can be taken into account, filtering the algorithmically-generated results based on specific variables.
This will not only have a significant effect in cutting down clinical trial time, but also help each patient to find their ideal medication more quickly, and without adverse effects from ineffective therapies.
Bearing these differences in mind, sufficiently programmed AI will also be able to deduce individual dose-sizes, using genetic and genomic data, sensors and smart devices to devise the particular needs of each patient.
An efficient future
While advances such as these often raise the question of whether robots are due to ‘replace’ humans, scientists will maintain their vital roles in drug discovery. Drug development AI is simply a sophisticated tool, used to supplement skills and give researchers more time to focus on patient-centric work.
With companies like BenevolentBio and AstraZeneca developing their algorithms to be faster and more accurate, the future is very bright indeed for drug development AI - and for the researchers that work with it.