/>iThe integration of artificial intelligence (AI) and machine learning (ML) into the life sciences field has created exciting new opportunities for advancements in diagnostics, therapeutics, and personalized medicine. However, obtaining patent protection for AI/ML-based inventions in life sciences can be difficult, particularly due to the challenges posed by US Patent & Trademark Office (USPTO) rejections under 35 U.S.C. § 101, which governs patentable subject matter. Patent applications directed to inventions at the intersection of AI/ML and life sciences are frequently rejected at the patent office under subject matter ineligibility grounds – more frequently than other life sciences inventions.
Patent applicants in this area must navigate the heightened scrutiny applied by patent examiners, who may issue rejections on dual grounds: (1) that claims are directed to an abstract idea, and (2) that claims are directed to a natural phenomenon. Life sciences enterprises are often familiar with Mayo Collaborative Services v. Prometheus Laboratories (2012), in which the Supreme Court ruled that natural laws, such as drug dosage correlations, are not patentable unless supplemented by additional, inventive steps. Meanwhile, in Alice Corp. v. CLS Bank International (2014), the Court held that abstract ideas such as those represented in computer-related inventions must include a significant inventive concept to be patent eligible under § 101.
Common Subject Matter Eligibility Concerns for AI Life Science Inventions
When life sciences companies utilizing AI/ML technologies pursue patent protection, it is common for their patent applications to face both types of subject matter eligibility rejections under § 101, particularly when the claims are viewed as directed to either abstract ideas or natural laws – or both.
Abstract Ideas
Inventions that incorporate AI/ML models often center around the use of data processing techniques, such as predictive modeling or the analysis of biological data. While these are powerful tools for advancing research and clinical applications, claims directed to these techniques are at risk of being categorized as directed to an abstract idea without significantly more technical elements necessary to be patent eligible. This risk is higher if a patent application’s claims are framed too broadly or focus only on general computational steps without clearly establishing a concrete, technical application. For example, claims that merely describe steps like "receiving data," "analyzing data," and "generating a prediction"—without grounding the process in a specific technical application such as within a specific biological or clinical workflow—risk being rejected under these grounds.
This is a common challenge for life sciences innovations that use AI/ML for purposes like disease prediction or biomarker discovery. Additionally, AI/ML technologies applied to image processing, such as radiological imaging, play a critical role in medical diagnostics. Without reciting an explicit technical improvement or enhancement to existing AI/ML techniques, or otherwise integrating the inventive concept into a practical application, claims to such inventions may not meet the threshold for patentability under § 101.
Natural Laws and Phenomena
Another frequent basis for rejection is raised against claims directed to AI/ML models that work with biological data, such as genetic correlations or physiological responses. In particular, when claims rely on the detection or analysis of natural relationships (e.g., the correlation between gene mutations and disease risk), they can be rejected under § 101 for reflecting a natural law without a sufficient technological improvement. Simply applying AI/ML to recognize or model these biological phenomena may not be enough; the claims must reflect a transformative application that goes beyond mere observation of natural laws. For example, an AI system that identifies genetic mutations linked to cancer may face rejection if the invention is viewed by the patent office as merely reflecting a known biological relationship.
Dual Rejection Scenarios – The “Double Whammy”
In some cases, AI/ML-assisted life sciences inventions encounter dual rejections, where claims are simultaneously deemed ineligible due to being directed both to abstract ideas and to natural phenomena. For instance, a patent claim that describes a machine learning model analyzing biological data might face rejection as directed to an abstract idea due to the computational focus while also being rejected for reflecting a natural law if it only observes natural biological relationships without representing a technological improvement or novel clinical application.
Examples of Rejections for AI/ML-assisted Life Science Inventions
Consider a claim directed at a machine learning model that analyzes genomic sequences to predict disease susceptibility. If the claim simply recites data processing steps—such as inputting genomic data, processing the data, and producing an outcome—it could be rejected as being directed to an abstract idea. This type of rejection is more likely if the claim does not connect the use of an ML model with a specific technical improvement represented by the invention.
Similarly, a claim involving AI technology to detect correlations between genetic mutations and cancer risk could be rejected under § 101 if it only describes the natural law—the relationship between mutations and disease risk—without presenting a transformative application, such as utilizing the AI model to guide therapeutic decisions in real-time, which might represent a practical and inventive application representing significantly more than simply an underlying abstract idea.
Strategies to Obtain Patent Eligible AI/ML Claims
Focus on Specific Technical Solutions
One of the most effective strategies for avoiding § 101 issues is to tie the AI/ML process to a specific technical solution within the life sciences context. Instead of merely claiming the computational steps or data analysis, the patent application should focus on how the AI/ML system enhances a particular biological or clinical process. For example, if the AI model is used to predict disease risk based on genomic data, the claims should emphasize any novel preprocessing techniques, feature extraction methods, or advanced machine learning architectures that contribute to faster diagnostics, more accurate predictions, or improved clinical outcomes.
This approach aligns with the Supreme Court’s guidance in Mayo by ensuring that claims involving natural phenomena, such as genetic correlations, include additional inventive steps that transform the natural law into a patent-eligible application. By focusing on how the AI/ML process improves the life sciences workflow—whether by reducing error rates, increasing efficiency, or providing more accurate diagnoses—the claims demonstrate a concrete, technical benefit that addresses both the Mayo and Alice issues.
Incorporate Real-World Transformations
Another strategy for avoiding § 101 issues is to incorporate real-world transformations or hardware integration into the claims. By showing how the AI/ML system is connected to physical devices or real-world processes, the claims can move beyond abstract ideas and natural phenomena. For instance, AI-driven diagnostic tools could interact with medical devices (including wearable devices) to process biological samples, or AI models could adjust therapeutic dosages in real-time based on patient data.
Linking the AI process to tangible outcomes, such as physical transformations or hardware interaction, strengthens the case for a claim’s eligibility. This approach directly addresses the Mayo requirement by showing how the AI system translates natural biological phenomena into a practical application. Likewise, it satisfies the Alice framework by introducing a significant inventive concept through real-world integration.
In drafting the claims, care should be taken to avoid divided infringement between different third parties that may eviscerate the strong protection that well-written claims can offer to patentees, especially now where there are many thousands of patent applications already filed with the USPTO.
Go Beyond Generic Data Processing
Finally, it is crucial to go beyond generic data processing in the claims. AI/ML systems in life sciences should be characterized in a way that highlights how they uniquely process biological data, offering specific benefits that set them apart from standard computational models. For example, the AI model may include innovative error correction mechanisms to handle noisy biological data, or it may dynamically adapt to real-time patient inputs to enhance predictive accuracy.
By emphasizing these unique data processing features, the claims move beyond generic data manipulation, making the invention more likely to be patent-eligible. This strategy addresses the dual concerns of Mayo and Alice, ensuring that the AI/ML system provides both a transformative application of natural laws and a significant inventive concept that distinguishes it from abstract ideas.
Conclusion
For life sciences companies developing AI/ML-assisted technologies, successfully navigating § 101 rejections requires a strategic focus on specific technical solutions, real-world applications, and innovative data processing methods. Beyond obtaining patents, enforcement under Section 101 raises additional considerations. Patent holders should be aware of potential attacks on validity during enforcement, particularly in litigation. Future articles in this series will delve into further strategies for successfully enforcing AI/ML life sciences patents. By partnering with sophisticated patent counsel to develop claims that highlight the tangible benefits and transformative use of AI/ML in biological and clinical contexts, life sciences companies can strengthen their patent applications and increase the likelihood of securing patent protection. As AI/ML continues to revolutionize the life sciences, robust patent strategies are critical to maintaining a competitive edge in this rapidly evolving industry.
