Artificial Intelligence in Drug Discovery

Written by: Mark Titleman

The importance of artificial intelligence in drug discovery cannot be overstated. The IBM Watson

supercomputer, for example, is a well-known instance of AI that has enabled the creation of

personalized treatment plans for cancer by correlating patient data with a vast database. These

treatment-centred correlations can be used for drug discovery as well as personalized treatment, and AI

has finally reached a developmental stage for more direct discovery. By processing and analyzing data

from gene expression profiles, pharmacology, chemical synthesis, drug design and drug screening, AI has

expedited the entire drug development process. One example is the recent work done by the

biotechnology company Berg in the design of a new cancer drug BPM31510 for pancreatic cancer

(currently in a phase II clinical trial). AI has become increasingly useful not only in medical treatment and

drug discovery overall but at every stage of drug development, from identifying potential drug

candidates to predicting their effectiveness and safety.

(Paul et al.)

Traditional drug discovery is a time consuming and expensive process that lasts many years and

requires significant multimillion-dollar investments. Rising attrition rates and extended clinical trial

duration have added to these inefficiencies, on top of an already 90% failure rate of potential drug

candidates to advance to phase-I clinical trial. Artificial intelligence is by far the most capable resource

for solving many of the inefficiencies endemic to the traditional process of drug development – a

process long burdened and stilted by incredible cost and precautionary exigencies. Several

biopharmaceutical companies have teamed up with IT companies for the express purpose of developing

AI platforms for drug discovery in areas ranging from immuno-oncology to cardiovascular diseases.

Many machine learning algorithms have been utilized in the pursuit of a more expedient process

for drug development. Bayesian network (BN), deep learning (DL), random forest (RF), clustering, and

support vector machine (SVM) are all used at different stages for different purposes. BNs predict toxicity

or bioactivity and patient response. DL models perform imaging, virtual screening, and bioactivity

predictions. RF models are used for target identification and selecting features for model building, while

clustering identifies patterns or relationships within the data. SVM classifies massive and multifaceted

data into categories to make further predictions. These techniques process data in many ways for

analysis and subsequent use in drug discovery and development.

Trained on large existing datasets that incorporate omics data, expression and phenotypic data,

and disease associations – as well as the mechanisms, tools, and outcomes of research – artificial

intelligence expedites the most complex task of protein and gene target identification. Drug efficacy is

also tested without the time and expense of traditional chemistry through computer-based molecular

simulations, predictions of chemical characteristics, and drug prioritization. AI looks for similarities

among compounds or project their toxicity based on input features. The recent Tox21 Data Challenge

evaluated several computational methods to forecast the toxicity of 12707 compounds and highlighted

a promising ML algorithm named DeepTox.

Chemical synthesis can be ameliorated, too, with novel synthesis pathways, existing compound

modifications, and complete de novo drug designs and their novelties being proposed. It is not out of

the question that fully automated end-to-end drug discovery will become a reality within one to two

decades. Even company response to market supply and demand may become automated; E-VAI is a

decision-making platform for making analytical roadmaps based on competitors, key stakeholders, and

currently held market share to predict crucial drivers in sales of pharmaceuticals.

With these new uses of AI in the context of drug discovery, drugs and treatments of all types

have been found for many diseases, including cancer, Alzheimer’s disease, and COVID-19, as the vastly

improved efficiency in developmental pipeline continues the never-ending medical search for cures.

Decrease in costs and appropriate scaling will eventually lead to tailored and affordable treatments for

the general population with well understood guidelines. This may seem like an impossible feat.

However, artificial intelligence in medicine will ultimately achieve its greatest benefits to humanity

through drug development.

References

Blanco-Gonzalez, Alexandre, Alfonso Cabezon, Alejandro Seco-Gonzalez, Daniel Conde-Torres, Paula

Antelo-Riveiro, Angel Pineiro, and Rebeca Garcia-Fandino. "The role of AI in drug discovery:

challenges, opportunities, and strategies." Pharmaceuticals 16, no. 6 (2023): 891.

Chun, Matthew. 2023. “How Artificial Intelligence Is Revolutionizing Drug Discovery.” Bill of Health.

March 20, 2023. https://blog.petrieflom.law.harvard.edu/2023/03/20/how-artificial-

intelligence-is-revolutionizing-drug-discovery/.

“New AI Tool for Rapid and Cost-Effective Drug Discovery.” 2024. Monash University. June 17, 2024.

https://www.monash.edu/news/articles/new-ai-tool-for-rapid-and-cost-effective-drug-

discovery.

‌ Mak, Kit-Kay, Yi-Hang Wong, and Mallikarjuna Rao Pichika. "Artificial intelligence in drug discovery and

development." Drug discovery and evaluation: safety and pharmacokinetic assays (2024): 1461-

1498.

Rehman, Ashfaq Ur, Mingyu Li, Binjian Wu, Yasir Ali, Salman Rasheed, Sana Shaheen, Xinyi Liu, Ray Luo,

and Jian Zhang. "Role of Artificial Intelligence in Revolutionizing Drug Discovery." Fundamental

Research (2024).