Artificial intelligence is revolutionising the pharmaceutical and biotechnology industries. Its unparalleled power to make predictions based on large data sets is advancing understanding of diseases, improving diagnostic speed and accuracy, accelerating drug discovery, and improving clinical trials.
AI's rapid growth makes it more important than ever for companies to safeguard their intellectual property.
AI as a diagnostic tool
AI is a powerful diagnostic tool. It can quickly analyze real-world data and detect subtle changes and patterns that humans may miss. In one study, researchers are testing whether AI can analyse MRI data to predict cancer patient outcomes better than well-trained radiologists. Similarly, a National Health Service-funded study is assessing AI's ability to analyze CT scans to diagnose blood clots, hoping it can discern information from the CT images that doctors might not see.
AI is not simply a tool for analyzing lab data. It can also develop new diagnostics, finding previously undetected correlations to diagnose or predict diseases.
AI and drug discovery
AI has tremendous power to identify, and even design, new therapeutic candidates using existing data. For example, AI can predict structure-function relationships for small-molecule drugs, identify new targets, and screen candidates by conducting molecular dynamics. Similarly, it can predict protein structure and function to identify new therapeutic candidates. These are just a few of the ways AI is improving drug discovery.
These capabilities translate directly to more therapies at lower costs. A recent analysis predicts that even modest AI-driven improvements in drug development could yield 50 additional therapies over 10 years, reflecting a $50 billion opportunity. AI may also reduce preclinical costs by 20-40% in some areas.
Companies are investing heavily to capture this opportunity. Third-party investment in AI-enabled drug discovery topped $5.2 billion in 2021. Meanwhile, pharmaceutical and biotechnology companies are investing billions developing internal AI capabilities and companies like Alphabet and Nvidia have expanded into drug research.
AI in clinical trials
AI presents an enormous opportunity to reduce the 10-15 years and $1 billion-plus needed to bring a new drug to market, much of which is due to costly clinical trials. Despite high costs, many clinical trials fail. One third of Phase II compounds never advance to Phase III and one third of Phase III compounds do not receive final regulatory approval.
AI can reduce clinical trial time and costs and increase success rates. It can improve participant selection by mining electronic health records and medical literature. AI can also improve patient monitoring using digital tracking, more accurately track disease progression using biomarkers, and improve analysis of safety and efficacy data. It can even find patterns in data that humans would miss, leading to new targets or treatments.
And, while randomised controlled trials will remain the gold standard in the near term for assessing safety and efficacy, AI is also enabling more targeted therapies and clinical trials that calibrate protocols based on patients-specific data.
Strategies for patenting artificial intelligence-based inventions
Effective claim drafting strategies can help overcome challenges in obtaining patents directed to AI-based inventions.
In the United States, the U.S. Supreme Court's Alice decision and subsequent § 101 cases have made it harder to obtain patents covering software- and computer-implemented inventions. Under Alice's two-part test, claims that are "directed to" a patent-ineligible abstract idea, law of nature, or natural phenomenon are only patent eligible if they contain an "inventive concept" sufficient to transform the underlying idea or law of nature.
Companies seeking to patent AI systems or software should draft claims that clearly reflect a technological improvement on conventional technology, with written description detailing that improvement. Reciting conventional computer technology, mere data gathering or processing steps, or restricting claims to a field of use (e.g. cancer diagnostics), are insufficient to provide an "inventive concept" that meets the Alice test.
Similarly, the Supreme Court's Mayo decision and its progeny made it harder to patent diagnostic methods. Companies seeking to patent AI-based diagnostics should draft claims that either reflect an improvement in the underlying AI or recite a treatment step based on the output of the diagnostic steps. Such treatment steps may provide an "inventive concept" that transforms the underlying data analysis into "significantly more" than any underlying abstract idea or natural phenomenon.
Different considerations apply in Europe for diagnostic methods claims. Article 53(c) of the European Patent Convention excludes diagnostic claims from patentability if they are carried out on the human body or include all the steps of collecting data, comparing it with standard values, identifying a symptom, and making a diagnosis. To avoid this exclusion, companies can draft claims directed to the underlying AI system or that avoid the step of providing a diagnosis or collecting a sample from the patient. Note, however, that the European Patent Office (EPO) has clarified that the diagnostic method exclusion cannot be avoided by omitting a step that is essential to practicing the disclosed invention.
Across jurisdictions, companies seeking to patent AI-based inventions must be careful to avoid inventorship issues. In most jurisdictions, including the EPO and U.S. Patent and Trademark Office, only humans can be inventors. Companies therefore should draft claims directed to contributions of human inventors and consider whether researchers involved with implementing the AI should also be listed.
Companies can also avoid rejections for insufficient disclosure by including enough description to show that the inventors possessed the full scope of the claimed invention and to permit a skilled artisan to practice the invention without undue experimentation.
Protecting valuable data and trade secrets
Finally, it is critical that pharmaceutical and biotechnology companies identify and protect their valuable data and trade secrets. Information ranging from protein libraries to methods of making specific compounds can be mined by AI systems to find new therapies or better understand diseases.
Protecting these data requires robust security measures, including labelling trade secrets, limiting access, training employees on secrecy, employing digital security measures, and ensuring third-party agreements require partners to use comparable secrecy measures.
Other considerations
AI's increasing prevalence has revealed instances of implicit gender- or race-based bias, which pharmaceutical and biotechnology companies must detect and eliminate. AI systems often struggle to generalise beyond the data used for training. Undetected racial or gender biases therefore could lead to misdiagnosis, undetected safety issues or contraindications, or inaccurate efficacy data for certain patients.
Conclusion
The greatest advances with AI in the pharmaceutical and biotechnology industries are yet to come. Companies that strategically protect their IP can position themselves for success and capitalise on the opportunities AI offers.
Originally Published by European Pharmaceutical Manufacturer
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
