As methods for generating in silico data continue to advance, there is an increasing interest in using such data to support the patentability of biotech inventions at the EPO. There are two main reasons for needing supporting data: to demonstrate sufficiency of disclosure (under Article 83 EPC) and to support inventive step (under Article 56 EPC). In the first article in this series, we focus on how in silico data may be relevant to sufficiency at the EPO.
What is in silico data?
The term “in silico” refers to data that have been generated using computational models of biological phenomena, rather than collected from “real-world” experiments, such as laboratory assays or clinical trials. This includes artificial intelligence-based approaches, as well as more traditional statistical and mathematical modelling.
The capability of computational models to generate such data has advanced rapidly over the past decade. One of the most notable examples of this is the Nobel Prize–winning molecular structure prediction program AlphaFold, which achieved unprecedented success in solving the “protein folding problem” (predicting the 3D structure of a protein from its sequence), and is now being applied to the prediction of proteins interactions with ligands, DNA, RNA and small molecules.
Given the rapid progress in computational modelling, it seems inevitable that more early-stage biological research will be conducted in silico, resulting in an increasing number of biological patent applications containing in silico data, and growing interest in the use of in silico data to support or challenge patentability, both before and after grant.
The EPO’s approach to in silico data
Despite the recent progress in computational modelling, the EPO’s approach to in silico data is relatively undeveloped at present. Case law addressing the use of in silico data in biotech patents is still extremely limited, particularly in the context of biologics such as antibodies. This may be, in part, due to the mismatch between the rates of technical progress and legal process: in many areas of biotech, computational modelling has only recently reached the point where it might be considered a viable and accessible alternative to data from wet experiments, while any issues arising in applications filed now are likely to take many years to reach the EPO's Boards of Appeal.
As a result, there is currently little guidance on how to make best use of in silico data. Some early indications as to how the EPO might approach in silico data come from EPO Board of Appeal decisions such as T 898/05, which indicates that in silico data can be suitable to demonstrate industrial applicability and “allow plausible conclusions to be made regarding the function of a product before it is actually tested”. Similarly, decision T 2517/19 held that experimental results reported for antibodies could be “credibly be extrapolated” to corresponding antibody-drug conjugates to support inventive step, based on an in silico study.
Although these decisions are fact-specific, they do indicate that the EPO is at least open to the use of in silico data to support patentability. While this is promising for applicants seeking a quick and cost-effective alternative to wet-lab experiments, caution remains essential: relying too heavily on predictive data that the EPO may in the future deem unreliable or unconvincing could prove extremely costly.
Sufficiency at the EPO: Putting the invention into practice
There is currently no clear guidance on how the EPO will assess in silico data in the context of sufficiency.
According to Article 83 EPC, to be sufficient, a patent application must include enough information to allow a skilled person to carry out the invention across the full scope of the claims without undue burden. If the claim recites a specific technical effect, the effect must also be achieved across the whole scope of the claim. These requirements must be met on filing. Although it is possible to provide additional data to support technical teachings already in the application, fundamental gaps in the information necessary to carry out the invention cannot be filled in later.
The bar for demonstrating sufficiency is typically low at the EPO. For product claims (e.g. directed to a therapeutic molecule), the skilled person need only be given enough information to make the product. This requirement is usually – although not always – straightforward for claims directed to a molecule that has been defined structurally, but becomes more complicated for molecules defined primarily in functional terms. The situation is similarly complex for medical use claims, as attaining the claimed therapeutic effect is also regarded as a functional feature of the claims. Consequently, for medical use claims to be sufficient, the therapeutic effect must be credible.
It seems possible that the EPO might accept in silico data in support of structural features or specific functional features such as antibody-epitope interactions that can be reliably predicted by structural information. Consistent with this, in decision T 835/21, the Board of Appeal concluded that “the preparation of a monoclonal antibody that binds to an epitope within a defined amino acid sequence … is a routine task for the skilled person and does not require any inventive activity”, suggesting that the absence of wet-lab data may not be a barrier to patentability for epitope claims - at least for linear epitopes which are easily isolated and reproduced for use in immunisation or phage display. This could be particularly advantageous for applicants, as determining the structure or epitope of antibodies and other binding molecules using traditional methods can be challenging, expensive, and time-consuming.
In silico data may also help applicants seeking broader claims by providing proof of principle that the skilled person can put the invention into effect across the whole scope of the claim, even if only specific embodiments have been exemplified with wet-lab data. For example, in silico data suggesting that specific modifications in an antibody sequence will not impact binding may support claims allowing for sequence flexibility.
In most cases, accurately predicting more complex functional features (such as binding affinity, internalisation, or neutralisation) remains beyond the scope of current computational models, making the EPO is unlikely to accept them on the basis of in silico data alone.
Using in silico data to support claims directed to medical use will also pose a challenge. In decision T 609/02, the Board of Appeal determined that for a medical use claim to be sufficiently disclosed, the application “must provide some information … to the avail that the claimed compound has a direct effect on a metabolic mechanism specifically involved in the disease”. The Board went on to suggest that “in vitro examples may be sufficient” if “they directly and unambiguously reflect the therapeutic application”. There is, in principle, no reason why in silico data could not also meet this standard. In fact, the concept of in silico clinical trials using “synthetic patients” or “digital twins” is generating significant interest in the pharmaceutical field. Nevertheless, this type of predictive modelling remains highly speculative – requiring simulation of multiple complex and interacting systems, potentially with limited data or an incomplete understanding of the underlying biology – and is therefore likely to be viewed with caution at the EPO.
Relying exclusively on in silico data is likely to be particularly risky for patents and patent applications expressing the technical effect in the claims, as failure to achieve a technical effect expressed in the claims results in insufficiency (as opposed to failure to achieve a technical effect not expressed in the claims, which leads to lack of inventive step). Insufficiency on filing cannot be remedied later, meaning that if the examiner is not convinced by the in silico data in the application, the likely outcome is refusal. It is therefore advisable to always include some real-world data if at all possible.
Opportunities for opponents
Many of the potential risks for applicants relying on in silico data translate into new opportunities for third parties looking to challenge the resulting patents. This is particularly the case for broad or speculative patents, where reliance on in silico data could be considered excessive or inappropriate.
The possibility of shifting the burden of proof in relation to sufficiency of disclosure represents another potential opportunity for opponents. In proceedings before the EPO, the burden of proof for an objection of insufficiency lies, as a rule, with the party raising the objection. The burden can be reversed if there are serious doubts as to the possibility of performing the invention. Historically, demonstrating such doubts has often involved carrying out costly wet-lab experiments, however, going forward, in silico simulations may present a faster and cheaper way to achieve the same result.
Conclusions
The EPO has yet to develop a standardised approach to the use of in silico data for supporting sufficiency of disclosure. Computational predictions may be acceptable in specific situations, for example, to provide evidence of structural features or certain binding characteristics. However, in silico data should not be seen as a direct substitute for experimental data.
For now, it appears the most prudent approach would be to use in silico data strategically – exploiting it for straightforward structural predictions and or well-characterised processes where acceptance is likely, particularly in situations where obtaining the corresponding real-world evidence is likely to be expensive and time consuming. This could reduce the experimental burden for initial filings, as well as enabling proprietors and opponents to provide evidence across a broader scope than might otherwise be feasible.
As computational methods continue to improve in accuracy and reliability, in silico data are likely to achieve wider acceptance and increased use in pre- and post-grant proceedings before the EPO. With the boundaries of biotechnology and computational science progressively converging, applicants would be well advised to strategically harness in silico data to support, expand upon, or reinforce real-world experimental validation – without, at least for now, losing sight of the enduring value of experimental evidence.
If you have any questions relating to in silico data and patentability at the EPO, please get in touch with Susan Hancock, or your usual Kilburn & Strode advisor.