AI-first drug discovery: Can robots help scientists discover new pharma drugs?

• Author:
• Author name

  Quantumrun Foresight
• August 22, 2022

Insight summary

High costs and failure rates in traditional drug development are pushing pharmaceutical companies to invest in artificial intelligence (AI) technologies to boost research efficiency and lower costs. AI is transforming the industry by rapidly identifying new drug targets and enabling personalized treatments. This shift towards AI is reshaping the pharmaceutical landscape, from changing job requirements for chemists to sparking debates on AI's intellectual property rights.

AI-first drug discovery context

The typical drug development project costs USD $2.6 billion. The pressure is high for scientists, as 9 out of 10 candidate therapies don’t reach regulatory approvals. As a result, pharmaceutical companies are aggressively investing in AI platforms during the 2020s to increase research efficacy while driving down costs. 

Different AI technologies are used in drug discovery, including machine learning (ML), natural language processing (NLP), and computer vision. ML analyzes data from various sources, including scientific literature, clinical trials, and patient records. This information can then be used to identify patterns that may suggest new drug targets or lead to the development of more effective treatments. NLP, a language-based predictive model, is utilized to mine data from scientific literature, which can highlight new ways that existing drugs could be developed. Finally, computer vision analyzes images of cells and tissues, which can identify changes associated with diseases.

An example of a pharma company that uses AI to develop new drugs is Pfizer, which utilizes IBM Watson, an ML system that can extensively research immuno-oncology medicines. Meanwhile, France-based Sanofi has partnered with UK startup Exscientia to create an AI platform to look for metabolic-disease therapies. Swiss company Roche subsidiary Genentech is utilizing an AI system from US-based GNS Healthcare to lead the search for cancer treatments. In China, biotech startup Meta Pharmaceuticals secured a USD $15-million seed funding to develop autoimmune disease treatments using AI. The company was incubated by another AI-assisted drug discovery company, Xtalpi.

Disruptive impact

Perhaps the most practical application of AI-first drug discovery was the development of the first therapeutic drug for COVID-19, an antiviral medication called Remdesivir. The drug was initially identified as a possible treatment for the virus by researchers at Gilead Sciences, a biotechnology company in California, using AI. The company used an algorithm to analyze data from the GenBank database, which contains information on all publicly available DNA sequences.

This algorithm identified two possible candidates, which Gilead Sciences synthesized and tested against the COVID-19 virus in a lab dish. Both candidates were found to be effective against the virus. One of these candidates was then selected for further development and testing in animals and humans. Remdesivir was ultimately found to be safe and effective, and was approved for use by the US Food and Drug Administration (FDA).

Since then, companies and organizations have collaborated to find more COVID-19 treatments using AI systems. In 2021, 10 companies banded together to create IMPECCABLE (Integrated Modeling PipelinE for COVID Cure by Assessing Better Leads). These organizations include Rutgers University, University College London, the US Department of Energy, Leibniz Supercomputing Center, and NVIDIA Corporation.

The project is an AI simulation pipeline that promises to fast-track the screening of potential COVID-19 drug candidates 50,000 times faster than current methods. IMPECCABLE combines various data processing, physics-based modeling and simulation, and ML technologies to create an AI that uses patterns in data to build predictive models. Unlike the typical method, where scientists have to think carefully and develop molecules based on their knowledge, this pipeline allows researchers to automatically screen huge numbers of chemicals, dramatically increasing the probability of finding a likely candidate.

Implications of AI-first drug discovery

Wider implications of industry adoption of AI-first drug discovery methodologies may include: 

• AI platforms assuming tasks traditionally handled by early-career chemists, necessitating these professionals to acquire new skills or shift career paths.


• Large pharmaceutical companies employing robotic scientists for scouring extensive genetic, disease, and treatment data, accelerating therapy development.


• A surge in partnerships between biotech startups and established pharma firms for AI-assisted drug discovery, attracting more investment from healthcare entities.


• The facilitation of tailored medical treatments for individuals with unique biological characteristics, especially those with uncommon autoimmune disorders.


• Intensified regulatory discussions on AI's intellectual property rights in drug discoveries and accountability for AI-related errors in the pharmaceutical sector.


• The healthcare industry experiencing significant cost reductions in drug development, allowing for more affordable medication prices for consumers.


• Employment dynamics in the pharmaceutical sector shifting, with an emphasis on data science and AI expertise over traditional pharmaceutical knowledge.


• Potential for improved global health outcomes due to faster and more efficient drug discovery processes, particularly in developing countries.


• Governments possibly enacting policies to ensure equitable access to AI-discovered medications, preventing monopolies and fostering wider health benefits.


• Environmental impacts lessening as AI-driven drug discovery reduces the need for resource-intensive laboratory experiments and trials.

Questions to consider

• How else do you think AI-first drug discovery will change healthcare?


• What can governments do to regulate AI-first drug developments, particularly pricing and accessibility?