FDA Accelerating the use of AI-Assisted Scientific Review

Nitrosamines in the News
1

Is your Nitrosamines risk assessment AI-ready?

https://content.govdelivery.com/accounts/USFDA/bulletins/3df85e7

3 Likes
2

Is AI ready for nitrosamines?

While steps to accelerate the approval of new drugs to market in general is to be applauded I can see risks with this.

It is still possible to get the wrong answers from LLM systems to relatively simple questions around the FDA limits of small nitrosamines. (I was playing around - seeing how easy it would be to get AI to do the start of a review of a product).

I hope that I am not being presumptive that there will still be expert, human oversight on the reviews that come out of any AI system.

4 Likes
3

If AI applied to nitrosamine topics will create false citations as already happened in legal questions, I will take the pop-corn and I will see the movie…
What Happened to the Lawyer Who Used ChatGPT? Lessons to Learn

5 Likes
4

I asked ChatGPT about using AI for nitrosamine risk assessment. Do you believe him? I’m worried that one day he’ll start telling me how to choose profitable stocks even though I haven’t asked. :grin:

Q. The US Food and Drug Administration (FDA) announced plans to implement AI tools throughout the agency to assist in conducting scientific reviews following the successful completion of a pilot program incorporating generative AI into reviews. How useful is AI in the risk assessment of nitrosamines? Please answer based on the risks, time, and cost benefits of using AI.

A. AI can be highly useful in the risk assessment of nitrosamines—a class of probable human carcinogens—especially when evaluated through the lenses of risk mitigation, time efficiency, and cost benefits. Here’s how AI contributes across these dimensions:

1. Risk Mitigation

  • Advanced pattern recognition: AI can detect structural alerts in chemical compounds that may lead to nitrosamine formation, even at trace levels. This helps in proactively flagging high-risk substances.
  • Predictive toxicology: Using machine learning (ML) models, AI can predict the carcinogenic potential of novel nitrosamine impurities based on known toxicological data, helping prioritize compounds for deeper analysis.
  • Structure-activity relationship (SAR) modeling: AI tools can analyze large datasets of nitrosamine structures to predict their DNA-reactivity and mutagenicity with higher precision than manual assessments.

2. Time Efficiency

  • Rapid screening: AI models can process and analyze thousands of compounds in hours—tasks that would take human reviewers weeks or months.
  • Automated documentation: Natural Language Processing (NLP) tools can help automate literature reviews, extract relevant data from scientific publications, and assist in drafting sections of regulatory submissions.
  • Real-time updates: AI systems can stay current with evolving scientific knowledge (e.g., new nitrosamine analogs) and adjust models accordingly without requiring complete re-analysis.

3. Cost Benefits

  • Reduced manpower needs: Automating routine tasks means fewer human resources are needed for initial assessments, freeing experts to focus on complex decision-making.
  • Minimized experimental costs: AI predictions can reduce the need for extensive in vitro or in vivo testing by identifying low-risk compounds early.
  • Early detection = fewer recalls: By identifying nitrosamine risks in early development stages, AI can help avoid costly drug recalls and regulatory penalties.

Conclusion

AI adds significant value in nitrosamine risk assessment by enhancing accuracy, reducing evaluation time, and lowering operational costs. While it should not replace expert judgment, AI serves as a powerful decision-support tool that complements and accelerates regulatory science. The FDA’s interest in expanding AI use reflects this strategic benefit.

Would you like a brief example of an AI model used for nitrosamine prediction (like QSAR or Deep Learning tools)?

2 Likes