'Genus' claims - breadth vs enablement

Patents are a legally-enforceable 'bargain' between inventors (and their funders) and the public: time-limited exclusivity rights for an innovation in exchange for public disclosure and ultimately the public benefit of such innovation. As part of a fair exchange, the patent holder must enable those working in the field of the invention to 'make and use' the invention for which exclusive rights are sought.

Many biotech and pharma patents include so-called 'genus' claims directed to a broad class of molecules defined around a limited number of working examples. Typically, the number of molecules/embodiments covered by such genus claims far exceeds the number that were made and tested or could ever be made and/or tested in practice. On the other hand, a claim limited to only the embodiments actually tested would, in many cases, provide an ineffective commercial barrier to the patentee’s competitors who could and likely would quickly design around the few claimed embodiments to exploit the patentee’s disclosed invention while avoiding infringement of the patentee’s claims. The tension between these considerations is a continuing theme in patent law, which is well illustrated by the recently decided Amgen v Sanofi case at the U.S. Supreme Court ('Amgen'), which suggests heightened scrutiny of genus claims can be expected during examination and in the courts. This is an especially important consideration for antibodies to new therapeutic targets.

Given the potentially mind-boggling number of possibilities typically encompassed by the broader claims in many biotech patents covering classes of therapeutic agents, should biotech and pharma innovators (and patent practitioners) despair of ever obtaining acceptably broad claims that are considered enabled? Perhaps not.

Leveraging AI for claim breadth

The rapid progress of AI-driven methods to rapidly and ever more accurately predict the structure and function of antibodies and other proteins based on their sequences promises a paradigm shift that will allow very efficient 'mining' of vast sequence spaces to yield genuses of in silico-vetted sequences. In effect, the strategic application of AI-driven platforms is likely to provide a far more efficient path to obtaining useful antibody and protein variants than afforded by conventional trial and error/conservative substitution approaches. Though obviously much smaller than a genus covering all possible sequence combinations, such “AI-enabled” genuses are likely broad enough to present an effective barrier to design-around by competitors trying to develop antibodies to the same target epitope. Importantly, as discussed below, claims to such in silico-vetted genuses will likely be increasingly accepted by patent offices and courts as enabled across their full scope and not, as currently cautioned by the U.S. Supreme Court, a mere 'hunting license'.

AI – a potential 'game changer' for enabling biotech therapeutic genus claims

Increasingly sophisticated developments in the application of AI-deep learning neural networks (DLNNs) to infer protein sequence-structure/function relationships is rapidly driving the evolution of a new paradigm to support enablement of sequence/structure-based genus claims. This is being accelerated by the availability of rapidly expanding protein sequence and structure databases to train DLNNs. Among the notable milestones in the development of these AI platforms is the arrival of DeepMind’s AlphaFold and AlphaFold21 and more recently, an antibody-focused platform, ABody Builder22.

Machine Learning approaches have been used to efficiently model 3D-antibody-antigen binding interactions of 6.9 million heavy chain CDR sequences to 159 antigens (1 billion antibody-antigen binding pairs) as demonstrated for the Absolut! simulation framework3. Importantly, these tools are being combined with methods to optimise searching in otherwise hopelessly large sequence spaces, e.g., using techniques such as Bayesian Optimisation to efficiently hone in on sequences that match or exceed the binding affinity of a starting antibody4.

Another very promising approach to sequence diversification and optimisation of antibodies is to generate an initial set of antibody (single mutations) to generate a training set of binding and non-binding sequences to develop a DLNN able to predict specific antigen-binding antibody sequences with high accuracy5.

Potential implications for biotech patent protection

While it is still too early to appreciate the full implications of AI-supported platforms to obtain genus claim coverage for antibodies and proteins, there is little doubt that such tools can be used to very great effect to leverage a small set of starting sequences/structures to define large, AI-vetted, genuses of functional variants.

A particularly powerful development is the beginning of 'self-driving laboratories' in which a ML-based platform generates protein sequence variants in silico and sends them directly to a fully automated robotic experimental platform to generate and experimentally validate the variant sequence protein. The outcome of such automated assays in turn allows the ML-based system to refine its protein designs to improve desired functional features. This 'virtuous cycle' can be expected to vastly increase the speed of protein engineering and enable sufficiently broad claims covering antibodies to new therapeutic targets/epitopes and proteins.

Key insights

  • Innovators seeking to maximise breadth around antibody/protein-sequence based claims should consider the use of AI platforms to generate a very broad and diverse set of sequences 'templated' around a handful of initially discovered sequences.
  • The increasing automation of functional assays for antibodies and proteins, especially when coupled to AI-based platforms to refine sequence generation will increasingly support the enablement of protein/antibody genus claims across their full scope.
  • The increasing sophistication of AI-protein design/validation platforms can be expected to be exploited by patentees to enable broader claims and their competitors to design around claims.

As AI transforms biotech, join us in shaping the future of your patents. If you are interested to find out more on navigating nuances, seizing opportunities, and keeping your IP at the forefront of this dynamic convergence, contact us for further information.



Jumper et al., (2021), “Highly accurate protein structure prediction with AlphaFold,” Nature, 596:583-589;


Abanades et al., (2023), “ImmuneBuilder: Deep-Learning models for predicting the structures of immune proteins,” Communications biology, 6:574.


Robert et al., (2021), “Unconstrained generation of synthetic antibody-antigen structures to guide machine learning methodology for real-world antibody specificity prediction,” bioRxiv, p. 2021.


Akbar et al., (2023), “Toward real-world automated antibody design with combinatorial Bayesian optimisation,” Cell Reports Methods 3, 100374.


Mason et al., (2021), “Optimisation of therapeutic antibodies by predicting antigen specificity from antibody sequence via deep learning,” Nature Biomedical Engineering, 5:600-612.

Want to know more? Explore our AI hub

Discover more about the future of innovation with our 'Beyond Human Ingenuity Series'
Whether you're a seasoned professional or an inquisitive mind, this hub is your one-stop destination for all things AI-related.
Read more