The McKinsey Global Institute (MGI) has estimated that Generative AI could generate $60 billion to $110 billion annually in economic value for the pharmaceutical and medical-product industries. This is primarily because it can boost productivity by accelerating the process of identifying compounds for possible new drugs, speeding their development and approval, and improving their marketing [1].
Target Discovery, Grand Challenges and Costs
In the research paper “How to Improve R&D Productivity: The Pharmaceutical Industry’s Grand Challenge,” Stephen M. Paul and colleagues address the critical issue of declining research and development (R&D) productivity in the biopharmaceutical industry. The paper highlights the increasing costs associated with discovering and developing new medicines. The authors analyse industry-wide data to identify factors contributing to reduced R&D productivity and propose strategies to improve it. The challenges faced by the pharmaceutical industry include patent expirations, cost constraints in healthcare systems, and demanding regulatory requirements. The goal is to enhance the quality and number of innovative, cost-effective new medicines while managing sustainable R&D costs [2].
The cost of developing a drug (for a known target, with a new molecule) in 2010, i.e., traditional R&D, was 1.8 billion dollars. However, it’s essential to recognise that these costs can vary significantly based on therapeutic area, success rates, and development time.
According to one of the latest studies in 2022, the average cost of bringing a new drug to market is approximately US$2 billion. This comprehensive process spans 14 years and includes drug discovery, pre-clinical and clinical trials, regulatory filings, and post-marketing surveillance [3].
Costs have stayed mostly the same. Although traditional AI methods started impacting pharmaceutical drug development in the late 1990s and early 2000s, the applications still needed to be improved.
However, the AI techniques employed in this early period were limited by data scarcity and computing power compared to modern standards. The real transformative impact of traditional AI in drug discovery started in the 2010s with:
- Availability of big biological data (genomics, proteomics etc.)
- Advances in deep learning and neural network algorithms
- Increased computing power via GPUs, cloud computing, etc.
How Generative AI is Transforming Drug Development
Now, Generative AI is playing an increasingly important role in various stages of the drug development process, including drug target identification, lead compound screening, and optimisation. Here’s a brief description of how Generative AI is being utilised in drug development today:
- Drug Target Identification:
- Generative AI models can design and propose novel molecular structures for potential drugs.
- These models analyse chemical databases, predict molecular properties, and propose new compounds with desired characteristics (e.g., high binding affinity to target proteins and low toxicity).
- This greatly expedites the early stages of drug discovery.
- Lead Compound Screening:
- Generative AI assists in the virtual screening of large compound libraries.
- It predicts how different chemical compounds will interact with the target protein.
- Researchers can identify potential leads with high affinity and specificity for the target.
- Optimisation:
- In the optimisation phase, AI refines lead compounds into more effective and safer drugs.
- It predicts how modifications to chemical structures could improve binding affinity and selectivity and reduce toxicity.
- Overall, Generative AI accelerates drug development by providing accurate predictions that guide experimental efforts effectively.
This transformative technology holds immense promise for faster, more intelligent, and more efficient drug discovery processes [4].
The possibilities for Generative AI in pharmaceuticals and biotechnology are vast and ever-expanding. Emerging trends indicate a shift towards more advanced AI platforms (such as Insilico’s PandaOmics) that can analyse data and autonomously formulate and evaluate hypotheses. Forecasts envision Generative AI playing a pivotal role in personalised medicine, utilising their capabilities to custom-tailor treatments to each individual’s unique genetic makeup [5].
The Game-Changing Benefits of Generative AI
Since the last drug development study in 2022, generative AI has entered our ecosystem. According to David Staunton, Head of Transformation, Life Sciences Manufacturing at Cognizant, traditional pharmaceutical companies have been able to synthesise around 1,000 molecules a year for drug discovery purposes. However, with the advent of AI, they can now analyse an astounding 50 billion molecules a year.
The positive impact Generative AI will have on pharmaceutical manufacturing:
- Reduced development costs
- Increased success rates
- Increased speed to launch
- Improved communications with healthcare professionals
Hence, AI can help accelerate this production timeline by allowing us to analyse, understand and automatically produce reports [6].
Leading the Charge: Pharma Companies Leveraging Generative AI Today
Sanofi announced its ambition to be “the first pharma company powered by artificial intelligence at scale” in a press release in June 2023. They kept to their word, as Sanofi, Formation Bio, and OpenAI collaborated in May 2024 to build AI-powered software to accelerate drug development and bring new medicines to patients more efficiently. The three teams combine data, software, and tuned models to develop custom, purpose-built solutions for the drug development lifecycle. This represents the first collaboration of its kind within the pharma and life sciences industries.
Sanofi will leverage this partnership to provide access to proprietary data to develop AI models as it continues to become the first biopharma company powered by AI at scale. Many other pharmaceutical giants are following suit and fully embracing AI to accelerate their drug discovery process, as detailed in the Pharma AI readiness index [7].
These same pharmaceutical giants have struck over 100 deals worth billions with AI drug discovery start-ups over the past decade. Advances in deep learning, computing power, and biological knowledge have enabled AI models to tackle complex tasks like predicting protein structures and generating entirely new molecule designs.
Navigating the Hurdles: Future Challenges in Drug Development
However, there is still scepticism about whether AI can substantially improve the dismal 90% failure rate in clinical trials, where biological complexity reigns.
While AI can accelerate drug discovery, its impact may be in failing faster and more cheaply. AI models can incorporate data from failed drug candidates to improve predictions. However, experts caution against overhyping AI’s potential benefits, as key factors like human biology and disease evolution remain poorly understood. AI is a powerful tool, but an AI-discovered drug still requires extensive human expertise and trials to validate its safety and efficacy for patients. This is why ‘build fast, fail fast’ doesn’t necessarily work for AI/ML in Life Sciences.
Most drug development failures occur during Phase 2 clinical trials. These trials involve testing the drug candidate on a larger group of patients to assess its efficacy and safety. These trials help determine whether the initial target, pathway, or disease hypothesis holds true in human patients. It can take decades to identify a good target for a disease.
For example, we still don’t have a perfect therapeutic target for Alzheimer’s or a treatment strategy to prevent it or completely eradicate it (although Eli Lilly’s Donanemab/ Kisunla holds promise for the early phases of Alzheimer’s disease). Thus, if we identify a perfect target during the early stages of drug discovery, there’s a higher probability of passing stage 2 of clinical trials [2].
While AI has attracted substantial investment, it hasn’t yet delivered on the promise of significantly faster and cheaper drug development. Challenges include translating AI-derived findings into clinical successes. To validate AI-driven discoveries, wet-lab validation and deeper biological understanding are needed.
The excitement around AI in the pharmaceutical industry centres on using AI tools to identify new drugs by analysing publicly available datasets. However, AI models excel at identifying correlations rather than causation. The success of generative AI methods depends on accurate and complete datasets.
Validating pharmacological approaches in real biological samples is crucial to bridge the gap. AI can accelerate novel discoveries and provide insights for clinical trial success. While AI is powerful for pattern recognition, human expertise is still needed to tease apart correlations from true causal explanations behind the identified relationships. Domain experts are still required to interpret the AI’s correlations and distinguish which represent meaningful causal links versus coincidental patterns in the data [8].
Safeguarding the Future: Essential AI Policies and Potential Risks
Finally, In February 2023, the European Medicine Agency released the Clinical Trial Registry (CTR); however, the associated documents do not mention the use of AI in drug development, which experts have said needs to be addressed. Pharmaceutical companies must have a voice in drug development regulating artificial intelligence (AI). The Veeva R&D Summit highlights this topic, emphasising the need for industry collaboration and proactive engagement to shape AI policies. As AI plays a significant role in clinical studies, understanding its potential risks and benefits is crucial for the pharmaceutical sector [9].
Takeaway
Integrating generative AI in drug and target discovery is shaking up the pharmaceutical industry by significantly enhancing productivity and reducing costs. While traditional AI has gradually improved drug development since the 1990s, generative AI is now expediting various stages, such as drug target identification, lead compound screening, and optimisation. This has led to more efficient processes, with the potential to analyse billions of molecules and tailor personalised treatments.
Despite these advancements, the industry still needs to improve, including high failure rates in clinical trials and the need for extensive human expertise to validate AI-driven discoveries. The future of AI in pharmaceuticals also hinges on developing robust policies to mitigate risks and ensure ethical practices, underscoring the importance of regulatory frameworks and industry collaboration.
References
[1] Generative AI in the pharmaceutical industry: Moving from hype to reality
https://www.mckinsey.com/industries/life-sciences/our-insights/generative-ai-in-the-pharmaceutical-industry-moving-from-hype-to-reality
[2] How to improve R&D productivity: the pharmaceutical industry’s grand challenge
https://www.nature.com/articles/nrd3078
[3] The Process and Costs of Drug Development (2022)
https://ftloscience.com/process-costs-drug-development/
[4] Artificial Intelligence in Drug Discovery and Development
https://link.springer.com/referenceworkentry/10.1007/978-3-030-73317-9_92-1
[5] Validated AI-Driven Target Identification and Biomarker Discovery with Insilico Medicine’s PandaOmics
https://www.bio-itworld.com/pressreleases/2024/02/27/validated-ai-driven-target-identification-and-biomarker-discovery-with-insilico-medicine-s-pandaomics
[6] Traditional and generative AI: an insight into the new normal for pharma firms
www.businessnews.ie/manufacturing/traditional-and-generative-ai-an-insight-into-the-new-normal-for-pharma-firms
[7] Pharma AI Readiness Index: Who’s best-positioned for the AI boom?
https://www.cbinsights.com/research/ai-readiness-index-pharma/
[8] AI Isn’t the Magic Bullet to Simplify Drug Discovery
www.genengnews.com/bioperspectives/ai-isnt-the-magic-bullet-to-simplify-drug-discovery/
[9] Veeva R&D Summit: Pharma companies need a say in AI regulation
https://www.clinicaltrialsarena.com/news/veeva-summit-ai-regulation-pharma-needs-day/?cf-view