Sequence identities and functional definitions - where is the limit? (T 0137/24)

Biotechnology patents often rely on defining protein and nucleic acid features by their sequence identity. However, it is often not necessary to define the exact sequence, but instead to define the sequence according to a limit of sequence identity, e.g. a sequence having a 90% sequence identity to a specified sequence. Prosecuting such applications is often about finding the limit of sequence identity that the Examiner is prepared to accept. The recent decision in T 0137/24, on a cannabinoid-producing yeast, reinforces the case law that low sequence identities should be permissible when combined with suitable functional definitions.  

Legal background: Added matter and sequence identity

An important aspect of the EPO’s strict approach to added matter under Article 123(2) EPC are the provisions relating to the selection of features from multiple independent lists. The EPO takes a famously strict approach to added matter and takes a hard line against what it considers to be “cherry-picking” of features from separate embodiments to arrive at a previously undisclosed combination. However, a nuanced exception to the strict prohibition on selecting from multiple lists is the concept of a convergent list. The EPO considers that, in non-convergent lists, each element is distinct. As such, the selection and combination of elements from two non-convergent lists therefore singles out a distinct combination from among several distinct alternatives. By contrast, in a converging list each progressively narrower element is considered to fall under the scope of the preceding elements. A convergent list thus is considered to provide a ranking of increasingly preferred options which allows patentees to narrow the scope of their claims without generating undisclosed combinations.

Percentage selection

Of course, the principles of selection from lists are very relevant to the biotechnology field given how biological molecules are routinely claimed. To provide a reasonable scope of protection, patentees often define a nucleic acid or protein by a minimum percentage of nucleotide or amino acid sequence identity to the specific sequence identification number used to define the sequence in the description and sequence listing (SEQ ID NO:). Consequently, patent applications in this field typically include lists of increasing sequence identities, such as "from at least 80% up to 100%", alongside separate lists of various sequences. It is therefore key to consider whether selection of a specific sequence identity for a particular sequence can be considered a disclosed or undisclosed selection. 

Case background: Cannabis yeast

The case related to The Regents of the University of California patent EP 3615667 for microorganisms and methods for producing cannabinoids and cannabinoid derivatives. The patent specifically claimed a genetically modified yeast cell comprising inter alia heterologous nucleic acids integrated into a chromosome. These nucleic acids encoded specific enzymes that were defined in the claim by their amino acid sequence identity. The main issues of appeal were added matter and sufficiency, with both issues relating to the sequence identity limitations. 

Added matter of sequence identity selections 

The Opponent argued on appeal that the amended claims contained added matter. The Opponent particularly submitted that the combination of a yeast cell with a chromosome integration feature and a specific 90% amino acid sequence identity, as specified in the claim of the Main Request, was not disclosed in the application as filed and added matter.  

The original application claims 2 and 3 defined the sequence identity as being “at least 65%” to SEQ ID NO: 110 or SEQ ID NO: 100 (a notably low sequence identity). The patent was granted with a sequence identity of 85%, whilst Claim 1 on appeal was limited still further to a sequence identity of at least 90%.

The Opponent argued that increasing the percentage from 65% to 90% alongside the chromosomal integration and yeast host cell features of the claim constituted an unallowable selection from multiple independent lists. However, the Board of Appeal was not convinced by this argument. The Board of Appeal noted that claims 39 and 40 of the application as filed provided a direct basis for a yeast host cell. Regarding the sequence identity the Board of Appeal observed that the application disclosed a list of increasing amino acid sequence identities to SEQ ID NO: 100 or SEQ ID NO: 110. This list started from at least 65% up to 100%. 

The Board of Appeal held that lists of increasing amino acid sequence identities are convergent lists of preferred options. Unlike non-convergent lists, where elements are distinct, convergent lists represent progressively narrower versions of the same feature (IPKat). Additionally, applying the case law from T 2134/10, the Board of Appeal found that the selection of a higher degree of sequence identity for a functionally defined polypeptide merely narrows down the polypeptides falling within the claim. For the Board of Appeal, such a selection from a convergent list does not single out a particular molecule or confer new properties to this molecule. The Board of Appeal thus concluded that the combination of these features did not introduce added matter.

Sufficiency of disclosure for sequence identities

The Opponent also argued that the claims encompassed non-working embodiments because many polypeptide sequences falling under the claimed definition lacked enzymatic activity, such as truncated polypeptides used in the examples of the patent itself. The Opponent argued that an undue research project was required to identity functional polypeptides. However, the Board of Appeal was also not convinced by this argument. Importantly, the Board of Appeal noted that the claims functionally defined the polypeptides by their enzymatic activity meaning non-functional sequences did not fall within the scope of the claims. The Board of Appeal observed that generating variants and testing their activity was routine for a skilled person: 

"generating polypeptide variants as recited in the claims, such as polypeptides comprising an amino acid sequence that has at least 90% sequence identity to SEQ ID NO:110 or SEQ ID NO:100, and testing them for their enzymatic activity are disclosed in the patent [...] and only require the use of techniques that are routine for the skilled person working in the technical field of recombinant enzyme variants. No evidence to the contrary was submitted by the appellant." (r. 29, emphasis added)

The Board of Appeal thus concluded that the invention was sufficiently disclosed. This was despite the fact that the only data in the application did not contain data relating to sequences with varying % identities to the specified SEQ ID NOs. In view of the Board of Appeal finding on sufficiency and added matter, the decision under appeal was set aside and the case was remitted to the Opposition Division with the order to maintain the patent in amended form.

Final thoughts

For those prosecuting applications in the biotech field, this is a very welcome decision, given the prevalence of long lists of various sequence identities generally included in applications. 

The Board of Appeal’s decision on sufficiency and non-working embodiments is worth noting. In the present case, the patentee limited the claims to at least 85% sequence identity during prosecution, but not in response to a specific Examiner objection. In view of the Board of Appeal decision it is possible that a  broader sequence identity may have been considered allowed in view of the capabilities of the skilled person to screen out non-working embodiments. 

This Kat wonders where the Boards of Appeal would have considered the cut-off of working embodiments for sequence identity to be. If a skilled person can test sequences with 95% sequence identity for functional compliance, what about sequences with 70% sequence identity? What about sequences with 50% sequence identity? Where does the undue burden begin? Notably, the EPO Guidelines for Examination state with respect to antibody inventions that "[i]t is possible to claim an antibody characterised by the sequences of both variable domains or CDRs with less than 100% sequence identity when combined with a clear functional feature" (G-II-6.1.4). It is left open as to how much "less than 100%" is acceptable, taking account of novelty and inventive step requirements. However, it is clear from this decision that the presence of non-working embodiments falling under the scope of a sequence identity definition may be overcome with judicious functional definitions. 

As an aside, it is rare to find a patent relating to sequences that doesn't contain some form of sequence error. Often these errors are not found in the important sequences, although in some cases they very much are, with serious consequences for the Patentee (IPKat). Sequence identity language can be your saving grace in cases where the sequence in your patent is not strictly 100% identical to the sequence of your actual product. 

Further reading

• Does the requirement for a "technical contribution" in a novel selection survive? (if selecting from convergent lists) (T 1621/16) (Mar 2020)
• Alexion's Soliris sequence error saga (Alexion v Samsung Bioepis [2025] EWHC 1240) (June 2025)