How To Navigate Patent Eligibility Of AI-Driven Diagnostics | #itsecurity | #infosec | #hacking | #aihp

By David McCombs, Vincent Shier, Eugene Goryunov, Dina Blikshteyn, Jamie Raju, and Brooke Cohen; Haynes Boone

In an attempt to formalize the law on patent eligibility under 35 U.S.C. § 101, the U.S. Supreme Court issued two major decisions outlining a two-part test: Mayo v. Prometheus and Alice v. CLS Bank.¹ Instead of bringing clarity, however, Mayo and Alice spawned discord among various district courts and, as a result, increased the uncertainty in § 101 jurisprudence. The U.S. Supreme Court had the opportunity to clarify the state of the law in American Axle, but declined to do so, denying the certiorari petition.²

The uncertainty in patent eligibility has had a particularly chilling effect on patenting life sciences and computer software inventions. AI-driven innovations that lie at the intersection of these two fields face particular difficulty in satisfying the judicially created requirements under § 101.

The prevailing concern for many life sciences companies is that they give the quid, but never see the quo that the patent system is supposed to provide. Given the publication of patent applications and the current challenges of obtaining and defending the patent protection arising, in large part, from the patent eligibility challenges, many life sciences companies are considering forgoing patent protection altogether. This decision is short-sighted. First, AI is transforming the life sciences industry and companies should strive to invest in and protect these game-changing inventions. Second, there are effective patent-drafting and claim strategies that increase the likelihood of passing muster under § 101 during prosecution and, subsequently, withstanding patent challenges in district courts.

In an article for Drug Discovery Online,³ we discussed how practitioners can navigate the uncertain area of inventorship in the area of AI-assisted drug discovery. In this article, we focus on patent eligibility of AI programs for disease diagnosis.

Investing In AI Is Vital For Life Sciences Companies

Given AI’s massive potential, it is no surprise that life sciences companies have developed and deployed AI programs to drastically improve a wide variety of technologies relating to human health. AI takes large data sets and uses highly complex mathematical formulas and statistical methods to predict desired outcomes, increasing the speed and accuracy of data analysis. In fact, the goal of AI is to mimic human intelligence by employing human-like reasoning skills but tackling more complicated analyses and significantly larger data sets than a human can realistically handle. Unlike humans, AI does not get tired or distracted, and when trained properly it produces consistent, accurate outcomes.

AI has had a significant and revolutionary impact on drug development, as discussed in our previous article. For example, using thousands, if not millions, of data points about a target protein, protein-molecule interactions, disease characteristics, and/or toxicity, life sciences companies are successfully implementing AI programs to identify a small number of drug candidates for further assessment. By using AI, companies are saving considerable time and resources compared to the traditional brute-force approach. In fact, several drugs that were developed using AI programs are currently undergoing FDA trials.⁴

Beyond drug discovery, AI is also bringing transformative change to the area of disease diagnosis. As with AI-driven drug discovery, AI-driven diagnostics harness AI’s data analysis capabilities to diagnose patients quickly and accurately by, for example, identifying patterns across a multitude of data inputs including genetic markers, physiological traits, symptoms presented, and family history. In some instances, AI programs have already diagnosed certain diseases better than physicians.⁵

Even though AI-driven inventions are key to the future of the life sciences industry and personalized medicine, the vague, unpredictable world of patent eligibility may scare companies away from seeking patent protection, instead opting to maintain these potentially game-changing developments as trade secrets.

The § 101 Jurisprudence

Of course, to be patent eligible in U.S., the claimed invention must be drawn to a process, a machine, an article of manufacture, or a composition of matter; however, the U.S. Supreme Court has previously outlined certain judicial exceptions, including abstract ideas, natural phenomena, and products of nature. Where one of these judicial exceptions is implicated in a claim, the U.S. Supreme Court outlined a two-part test for patent eligibility under § 101 in Mayo, which was reaffirmed in Alice.  

Step One of this test involves analyzing whether the claims of the patent are directed to a judicially recognized exception.⁶ This process is anything but simple and often involves analogizing the claimed invention to the previously decided cases. If the claims involve an abstract idea, natural phenomenon, or a product of nature, like many AI-driven life sciences innovations, then the analysis proceeds to Step Two.⁷ Step Two asks whether the claims recite additional elements that contain an inventive concept beyond the judicial exception to satisfy §101.⁸

For a claim “to transform” a patent-ineligible exception into a patent-eligible invention in Step Two, the claims must do more than “simply state the [exception] while adding the words ‘apply it.’”⁹ The claims must explain how the invention, even though it uses a judicially recognized exception, improves concrete technology (i.e., how is the exception integrated into a practical application?).¹⁰ However, a significant impediment is that claims cannot be patent eligible simply by applying the exception to routine, conventional, or well-known technologies or by using routine, conventional steps.¹¹ Thus, even though the claims may be directed to ground-breaking discoveries, if the invention is a judicially recognized exception, the invention will not be patent eligible.¹²

Even if a patent is granted, there are no guarantees that the patent will survive, as patent eligibility can be challenged at any time during the life of the patent. When a patent is challenged in district court during an infringement case, judges determine whether the claims are patent eligible. In some cases, patents have been rejected under § 101 quite early in the litigation process. For example, in Health Discovery, claims were held patent ineligible in a § 12(b)(6) motion to dismiss.¹³ Others, like Mayo and Alice, take years and multiple appellate rounds before a final determination is rendered.

Unfortunately, the life sciences industry is in the eye of the patent eligibility storm. Judicially recognized exceptions are directly implicated in many life sciences innovations as these inventions necessarily relate to chemical and biological laws and, thus, incorporate natural phenomena. This is especially true for diagnostic methods, which require a recognition of a particular biological “truth” that correlates to a particular disease.

Further, AI relies upon algorithms that implement improved mathematical formulas or statistical concepts to better predict a certain phenomenon. As such, essentially all AI inventions involve abstract ideas. Thus, when putting together diagnostic methods and the use of AI, innovators have their backs against the wall. When an AI program performs diagnosis, the patent eligibility problem essentially amounts to applying a technology based on abstract ideas to a technology based on natural phenomenon.

How Can Practitioners Alleviate Patent Eligibility Issues?

Although courts have had difficulty making sense of § 101 law and applying it uniformly, patent practitioners can fortify patent applications to better weather patent eligibility challenges.

Practitioners should tie both the claims and the specification to improvements in a concrete technology. Rather than simply writing claims that focus on the judicially recognized exception, the claims should emphasize the application of the judicially recognized exception to a concrete technology. This process may include defining a specific system that incorporates the innovation related to the judicially recognized exception and explaining how the innovation improves the functioning of the system as a whole. Further, as the analysis is to focus on the “claim as a whole,” it is also important to integrate the judicial exception into a practical application and argue that the entire claim is patent eligible, even if the individual limitations are not.

Where possible, a practitioner can describe technical steps rather than rely on steps recited at a high level of generality. Claims that recite generic steps like obtaining a sample, analyzing a sample, correlating a result, storing data, retrieving data, or transmitting data are unlikely to overcome patent eligibility issues in front of the U.S. Patent Office, and even if they do, are difficult to defend if the patent is challenged in district courts. Instead, when using an AI program that performs a diagnosis or predicts a drug candidate, the claims should explain how the AI program works, how it encodes, decodes, and otherwise manipulates data, computer memory, etc., and how it arrives at a desired result. Although these types of claims are more difficult to craft and often require understanding of both life sciences and computer technologies, these claims are more likely to withstand patent eligibility scrutiny in front of the U.S. Patent Office and the district courts.

To provide context and support for the claims, practitioners should also include a discussion of the advantages of the innovation in the specification. This discussion should focus on explaining how the invention improves a specific technology and why these improvements are separate from the judicially recognized exception. Instead of identifying vague advantages to generic technologies, it is more effective to be as specific as possible. For example, a practitioner can overcome a patent eligibility hurdle at the U.S. Patent Office by explaining that an AI program that predicts drug candidates directly effects the speed, accuracy, and cost of drug discovery; what diseases or types of drugs the program predicts; or specific dosages of the drug that are prescribed during treatment. At the same time, these arguments may or may not get traction in district courts. In Health Discovery, for example, Judge Albright discounted the speed and accuracy benefits of the invention when he determined that the claims are not patent eligible.¹⁴ It is also important to focus on the basic and novel characteristics of the invention, whether it be through the use of a particular new step and/or element or whether it is a novel combination of steps and/or elements.

However, practitioners should be wary of clearing the § 101 hurdle at the expense of receiving a useful claim scope. Instead, practitioners must consider whether the claims that are patent eligible provide a reasonable scope of protection in view of the desired commercial application.

When drafting patent applications involving AI-driven diagnostics, practitioners can consider how to reframe the diagnostics claims into a type of claim that courts are more likely to find patent eligible. For example, courts have been more likely to allow a method of treatment claims.¹⁵ Thus, if the AI-driven diagnostic identifies a particular disease, the claims may include how the result of the diagnostic test can affect the treatment. In this case, practitioners could include specific treatment regimens involving, for example, specific doses of medication to be prescribed upon a specific result.¹⁶ However, a problem arises where the claims merge diagnostic steps with therapeutic steps to overcome a § 101 rejection. This type of claiming could inadvertently walk into a divided infringement situation where one party does the diagnosis using, for example, an AI program and then a physician (a second party) administers treatment. Following Akami,¹⁷ these claims are virtually impossible to enforce and would be less valuable.

Another option would be to write the diagnostic claims as a method of preparation claim.¹⁸ This option may not be viable for many technologies because it involves improvements to preparing a sample for testing, but practitioners should still be on the lookout for when it may be available. Again, such a claim may have limited value and may implicate divided infringement.

Although method of treatment and method of preparation claims are good options for making AI-driven diagnostic claims patent eligible, the most effective option may not include rebranding the claims. Instead, practitioners may have more success by focusing on how the AI program improves the functioning of the test itself.¹⁹ In some instances, the AI program improves the functioning of a computer by producing a better graphic display to improve the ease of making a diagnosis or by increasing the processing speed or accuracy of the program. For other inventions, the AI program may impact the physical testing of a patient by changing the way a test is performed. Where possible, practitioners should be on the lookout for these practical applications to include in the patent application and also in the claims.

Companies Should Continue To Invest In Patents For Their AI-Driven Innovations

Even though the task of overcoming patent eligibility requirements under § 101 seems daunting, there are ways to provide strong patent protection while minimizing the risk that the patent will be invalidated. The savvy patent practitioner can identify potential patent eligibility issues and take effective measures to address § 101 while obtaining robust claims. Thus, although the patent eligibility landscape is challenging and unpredictable, it is not entirely bleak and should not impact life sciences companies’ investments in AI-driven technologies or patents thereof.

This article reflects only the present personal considerations, opinions, and/or views of the authors, which should not be attributed to any of the authors’ current or prior law firm(s) or former or present clients.

References

1. Mayo Collaborative Servs. v. Prometheus Labs., Inc., 566 U.S. 66 (2012); Alice Corp. Pty. Ltd. v. CLS Bank Int’l, 573 U.S. 208 (2014).
2. American Axle & Manufacturing, Inc. v. Neapco Holdings LLC, 967 F.3d 1285 (Fed. Cir. 2020), cert. denied, — U.S. –, 2022 WL 2347622 (2022).
3. How To Navigate The Patenting Challenges Of AI-Assisted Drug Discovery, LIFE SCI. CONNECT, available at: https://www.pharmaceuticalonline.com/doc/how-to-navigate-the-patenting-challenges-of-ai-assisted-drug-discovery-0001 (June 17, 2022).
4. EXSCIENTIA, https://www.exscientia.ai/pipeline (last visited Feb. 28, 2022); RECURSION PHARMACEUTICALS, https://www.recursion.com/pipeline (last visited Feb. 28, 2022); BERG HEALTH, https://www.berghealth.com/research/healthcare-professionals/pipeline/bpm31510/ (last visited Feb. 28, 2022); BERG HEALTH, https://www.berghealth.com/research/healthcare-professionals/pipeline/bpm31543/ (last visited Feb. 28, 2022).
5. For instance, an AI program predicted breast cancer better than radiologists. Scott Mayer McKinney et al., International evaluation of an AI system for breast cancer screening, 577 Nature 89, 94 (2020).
6. Alice Corp., 573 U.S. at 208.
7. Id. at 208-209.
8. Id. at 209.
9. Mayo Collaborative Servs., 566 U.S. at 77-78
10. Id. at 79-80.
11. Id.
12. Ass’n for Molecular Pathology v. Myriad Genetics, Inc., 569 U.S. 576, 591 (2013); In re Bd. of Trs. of the Leland Stanford Junior Univ., 991 F.3d 1245, 1251 (Fed. Cir. 2021).
13. Health Discovery Corp. v. Intel Corp., 6:20-cv-666-ADA, 2021 U.S. Dist. LEXIS 245515 at *35-*36 (W.D. Tex. 2021).
14. Health Discovery, 6:20-cv-666-ADA, 2021 U.S. Dist. LEXIS 245515 at *32.
15. Vanda Pharms., Inc. v. West-Ward Pharms. Int’l Ltd., 887 F.3d 1117, 1134-1135 (Fed. Cir. 2018).
16. However, practitioners should be wary of including claim language directing physicians to not to give treatment based on the results of the diagnostic test. INO Therapeutics LLC v. Praxair Distrib., Inc., 782 Fed. Appx. 1001, 1007 (Fed. Cir. 2019).
17. Akamai Techs., Inc. v. Limelight Networks, Inc., 797 F.3d 1020 (Fed. Cir. 2015) (en banc).
18. Illumina, Inc. v. Ariosa Diagnostics, Inc., 967 F.3d 1319, 1325 (Fed. Cir. 2020).
19. XY, LLC v. Trans Ova Genetics, LC, 968 F.3d 1323, 1330-1331 (Fed. Cir. 2020); Enfish, LLC v. Microsoft Corp., 822 F.3d 1327, 1337 (Fed. Cir. 2016); See also CardioNet, LLC v. InfoBionic, Inc., 955 F.3d 1358, 1368 (Fed. Cir. 2020).