Recent advancements in computational modelling are changing the nature of biological research, particularly in fields such as drug discovery and antibody engineering. As discussed in the previous article in this series, this has led to a growing number of biotechnological patent applications incorporating or relying on in silico data. Yet, despite the growing interest in exploiting computational models for biotech research, the legal framework defining what constitutes suitable evidence for the patentability of biotech inventions has not kept pace. As a result, significant uncertainty exists for those seeking to rely on in silico data to support or challenge patentability.
This article explores the challenges and opportunities of using in silico data in relation to novelty and inventive step at the EPO.
Inventive step at the EPO: Establishing a technical advantage
For subject matter to be considered inventive at the EPO, it must demonstrate an improved or unexpected technical advantage over existing technology. For biological inventions, data is generally required to support the technical advantage relied upon.
Unlike the USPTO, the EPO does not apply the principle of structural non-obviousness, meaning a new molecule will not be considered inventive simply because it has a new structure. Instead, the difference in structure must give rise to a surprising or advantageous property, such as improved affinity, stability, or therapeutically relevant activity.
Data to support an inventive step at the EPO can be filed as part of the application itself, or can be submitted as support evidence after filing of the application (‘post-filed data’) during prosecution or opposition proceedings. There are no restrictions on the data that can be included on filing, but in order to rely on post-filed data for an inventive step, the technical effect demonstrated by the post-filed data must be “encompassed by the technical teaching” of the original application, and ”embodied by the same originally disclosed invention” (as confirmed in the Enlarged Board of Appeal decision G 2/21).
In silico data to support an inventive step: Challenges and opportunities
Establishing a “technical teaching” in the application as originally filed (in the sense of G 2/21) represents a particularly promising potential use for in silico data. If the EPO were to accept computational predictions as evidence of a technical teaching, this would ease the pressure on applicants to perform time-consuming or expensive experiments upfront, while keeping open the option of verifying any particularly relevant technical effects using real-world data as their research progresses.
In contrast, applicants seeking to rely exclusively on in silico data to meet the EPO’s inventive step standard are likely to face significant challenges. Currently, real-world data are generally seen as essential to validate technical effects that may be considered surprising or advantageous, and it is unlikely that this will change any time soon. Although computational modelling can be extremely powerful, it is not infallible; even for state-of-the-art programs such as AlphaFold, accuracy in predicting certain types of interactions remains low. Moreover, algorithmic errors often differ fundamentally from human errors and can be challenging to pick up, especially when they originate from the model's learned parameters rather than human input.
There is also a risk that in silico data may be detrimental if used prematurely. Applicants relying exclusively on in silico predictions may find the molecules they choose to cover in their initial filings subsequently fail validation, cannot be validated within the necessary timeframe, or require extensive optimisation to be clinically useful. The resulting filings are unlikely to provide adequate protection for any final commercial products, yet may still be prejudicial for novelty and inventive step against future applications. Publicising premature or failed predictions may also cast doubt on the validity of the methods used, leading to potential sufficiency issues for future applications relying on the same methods. Careful consideration of the reliability of the in silico data, the practicalities of validating the predictions made, and any subsequent optimisation needed to develop the claimed invention into a commercially viable product are therefore essential when deciding when to file.
The type of technical effect supported by in silico data is also significant. Accurate modelling of complete biological systems, such as living organisms, remains beyond the scope of current computational methods. In the context of biologics, in silico data may be considered reliable enough to provide evidence of a specific and well-defined molecular interaction (for example, binding of an antibody to an appropriate epitope). However, it is unlikely to be accepted as evidence of a desired therapeutic effect, where multiple other factors such as pharmacokinetics, immunogenicity and cross-reactivity can come into play. Nevertheless, computational modelling of biological systems remains an active area of research, and the EPO’s approach will inevitably evolve as technology progresses towards more comprehensive whole-system or whole-organism models.
As computational methods become more established, the bar for what is considered surprising or unexpected by the EPO may also shift. For example, under the EPO’s current approach, improved stability is considered a surprising technical effect for the purpose of assessing inventive step. Consequently, using techniques such as molecular dynamics to identify and stabilise weak spots in an antibody framework region is likely to produce antibodies which would be considered inventive due to their improved stability. However, it could also be argued that such improvements are simply the result of applying a known method (molecular dynamics) for a known purpose (improving stability). If the EPO were to endorse this line of reasoning, it may become significantly more challenging to obtain protection for engineered antibodies.
In silico data can also be helpful to establish novelty
In silico data may also be helpful in establishing novelty. Unlike inventive step, data are not typically required for the assessment of novelty at the EPO, however, there are situations where data become relevant. In particular, for certain functional claims, such as those defining an antibody by its epitope or by unusual parameters (parameters which are not routinely measured or reported in the literature), the EPO has taken the position that the responsibility for demonstrating that the feature provides a genuine distinction over the prior art lies with the applicant or proprietor.
Consequently, defining inventions using functional or non-standard features can be extremely challenging; even in cases where the feature in question does provide a genuine distinction over the prior art, generating the data to demonstrate this can be expensive and time-consuming. However, if the EPO were willing to accept in silico data to verify the absence of such properties in the prior art instead, the barriers to obtaining these types of claims might be significantly lowered. This may become particularly relevant for epitope claims, as recent progress in modelling epitope-paratope interactions could make in silico epitope screening a far more practical prospect than obtaining and screening physical antibodies in the near future.
Opportunities for opponents
In silico data also presents opportunities for those seeking to challenge patents at the EPO. When attacking patents relying on in silico data, many of the limitations discussed above can be translated into new avenues to challenge validity. For example, third parties seeking to challenge patents using in silico data to demonstrate a technical effect should consider raising questions about the reliability and transparency of computational models, particularly where full details of model architecture, training data, or validation methods are not disclosed, or where the alleged technical effect is complex, poorly characterised, or lacks a clear mechanistic explanation.
Alternatively, if the invention itself was developed using in silico optimisation approaches, it may be possible to argue that it represents the routine application of a known technique.
Opponents may also wish to make strategic use of their own in silico data in support of a novelty or inventive step attack. For example, in silico data can be used to strengthen an argument that a claim is unduly broad by predicting situations where the technical effect will not be achieved, or to demonstrate that certain optimisations are obvious by re-running simulations using common general knowledge to arrive at the same result. This could be particularly valuable in situations where the time or costs associated with generating equivalent in vivo or in vitro data are not feasible.
Conclusions
In silico data present both significant opportunities and considerable challenges for biotechnology patent practice at the EPO. The use of in silico data to support novelty and inventive step has the potential to ease the experimental burden prior to filing, support broader claim scope, as well as opening the door to new types of claims relying on functional features or technical effects that may have previously been deemed too burdensome to demonstrate.
The EPO's approach towards in silico data is likely to remain cautious and case-specific. Relying on in silico data to support novelty and inventive step is therefore not without risk, particularly for inventions at the boundary of current predictive capabilities, such as those relating to complex physiological effects or therapeutic efficacy. For now, the use of in silico data should be viewed as a supplement to, rather than a substitute for, experimental validation.
Third parties seeking to challenge patents should be aware of the potential pitfalls surrounding in silico data and look out for opportunities to leverage these effectively. Concerns regarding model transparency, lack of validation, and the appropriateness of computational evidence for demonstrating complex biological effects can be used to undermine inventive step, particularly where methodological details are lacking. Opponents should also consider generating their own in silico data to challenge technical disclosures that are speculative or obvious.
Success in this evolving landscape will depend on understanding the capabilities and limitations of current computational methods, and crafting patent strategies that appropriately balance the efficiency gains of in silico approaches with the EPO's patentability requirements. As computational methods continue to advance and scientific consensus develops around validation standards, the EPO's approach to in silico evidence will likely evolve accordingly.
If you have any questions relating to in silico data and patentability at the EPO, please get in touch with Susan Hancock, Amy Nick, or your usual Kilburn & Strode advisor.