2026 Week 5
For good or bad, it is now clear that inter partes preliminary injunction (PI) proceedings at the UPC end up being mini-trials of the substantive issues. In Guardant Health v. Sophia Genetics, Guardant requested a PI against Sophia based on three European patents. The UPC’s Paris Local Division stepped through the issues in detail and came to the view that, although Guardant had not delayed for too long before starting the PI proceedings, there were too many doubts about the validity or infringement of the asserted patents to grant the PI.
Background
Guardant requested the PI against Sophia’s liquid biopsy tests across the UPC Member States and beyond. Two of the asserted patents, EP 3591073 (“EP’073”) and EP 3443066 (“EP’066”), covered wet steps such as sample preparation and DNA sequencing. One, EP 3766986 (“EP’986”), relates exclusively to dry steps (analytics).
Sophia’s liquid biopsy tests analyse ‘cell‑free DNA’ (cfDNA) taken from a patient’s blood to detect and monitor genetic alterations associated with cancer. These mutations may be present at much lower levels than the unmutated sequences. Besides detecting variant (mutated) sequences, which might be present at very low levels, Sophia’s tests also establish a ‘consensus’ sequence, which is taken to represent the unmutated, original sequence.
The asserted patents refer to “collapsing” the sequences to form the consensus sequence. The patents describe two ways of doing this; either by taking the most frequent nucleotide at each position of the sequencing read, and more probabilistic approaches.
Probable added matter
The court decided that both EP’073 and EP’066 are more likely than not invalid on the grounds of added matter.
For EP’073, the court found that Guardant had “combinatorially” selected disparate features (cfDNA from bodily fluid; the quantity of ≤100 ng; enzymatic ligation; and non‑unique barcodes coupled with start/end sequence) without the originally filed application providing a clear pointer to the claimed combination. The court made a statement that is commonly found in the EPO’s case law on added matter:
“…the content of an application must not be considered to be a reservoir from which features pertaining to separate embodiments of the application could be combined in order to artificially create a particular embodiment.” (Paragraph 83 of the decision, in part.)
Given this EPO-style opinion on added matter, you might have thought that this would have been the EPO opposition division’s view of the validity of the same claim. But in fact the EPO opposition division had upheld EP’073 in slightly amended form, concluding that the claims upheld did not contain added matter. Indeed, Guardant’s defence in the UPC PI proceedings seemed to largely consist of simply referring to the OD decision in their favour.
Notably, after the oral hearing in the PI case on 12 December 2025, the EPO Board of Appeal issued its preliminary opinion in the opposition appeal. This clearly indicates the EPO Board of Appeal’s intention to overturn the OD decision and revoke the patent for added matter at the EPO appeal hearing scheduled for April 2026. Sophia requested that this preliminary opinion be admitted into the UPC PI proceedings on 16 December 2025 and the LD agreed.
For EP’066, the court’s added matter concerns arose from the selection of a list of four cancers recited in the granted claims, in contrast to the list of five cancers disclosed in the underlying patent application as originally filed, along with the addition of the term “one or more loci” in the granted claims, which was not present in the application as originally filed.
Question marks over infringement
The LD also raised significant questions over whether EP’073 and EP’986 were infringed.
For EP’073, claim 1 culminates in a step of “collapsing” the sequencing reads to form a consensus sequence. The LD accepted that the description of EP’073’s mentions two analysis paradigms—(i) base‑to‑base consensus and (ii) probabilistic approaches—but held that claim 1 only covers the base‑to‑base consensus method. The probabilistic approach, though discussed in the description, falls outside of the claims. The court accepted Sophia’s contention that the processing in its test is only probabilistic—not the base‑to‑base approach as claimed.
For EP’986, claim 1 defines:
“A computer implemented method comprising use of a computer database to identify one or more effective therapeutic interventions for a subject having cancer, wherein the computer database includes, for each of a plurality of subjects having cancer:
(i) tumor genomic testing data, including somatic alterations, collected at two or more time intervals per subject via serial biopsy of cell-free DNA;
(ii) one or more therapeutic interventions administered to each of the subjects at one or more times; and
(iii) efficacy of the therapeutic interventions…”.
To argue infringement, Guardant relied on Sofia’s product marketing and a press release that hinted at highlights the use of their test for “longitudinal monitoring” of a patient over time.
However, the court was not persuaded that Guardant has provided enough evidence for infringement, commenting:
“the accused test is meant to retrospectively analyse cancer treatment in clinical trials and to use this information to refine and optimise clinical trial design and improve patient recruitment for trials. The refining and supporting of clinical trial design is not the same as the identification of one or more effective therapeutic interventions.” (Paragraph 146 of the decision.)
No undue delay
Although their view on the substance of the case was that the PI should not be granted, the LD set out in detail their views on the timing of the filing of the PI and in particular whether this indicated that the patent proprietor had delayed too long. As we have seen in other cases (e.g. Cilag v. Rivolution, reviewed in UPC Weekly 2025 w36), this in itself can lead to the refusal of a PI.
The timeline of Guardant’s knowledge of the alleged infringement was disputed between the parties, with Sophia arguing that Guardant should have been aware of their products and systems in mid-2024. However, the LD decided that there was not enough evidence of this, and in particular whether the relevant employees of Guardant who might have been aware of Sophia’s activities in 2024 would have had the necessary IP knowledge or seniority to decide on a judicial action.
The LD therefore decided that the “trigger date” should be taken as 27 May 2025, the date of a warning letter sent by Guardant to Sophia concerning alleged infringement in the UK. Patent infringement proceedings were then started in the UK on 14 July 2025.
Guardant explained that they then investigated other countries in Europe and found possible infringements, and so started the UPC PI proceedings on 29 August 2025. The court took into account an argument from Guardant that technical analysis of the alleged infringement was not easy, and that further evidence was only obtained in July 2025.
Bearing in mind the front-loading of the UPC procedure, a claimant at the UPC needs to put in a lot of work to prepare the evidence and arguments for a PI application, which takes some time. The key observation from the court was this:
The Court considers that a three-month period (until the application dated 29 August 2025) constitutes a reasonable delay to prepare its action by gathering the necessary evidence, given that the case involves several patents and a complex and sophisticated technology. (Paragraph 36 of the decision.)
Every case is different, and so this 3 month data point can’t necessarily be transposed. But for future PI applications involving multiple patents, complex technology and factually difficult infringement analyses, it could easily be the new normal in view of the UPC’s expectation that the patentee will need time to prepare their complete case at the outset.
This blog was co-authored by Eliot Ward and Matthew Naylor.
Matthew Naylor
Matthew is a UPC Representative and European Patent Attorney. He is a Partner and Litigator at Mewburn Ellis. He handles patent and design work in the fields of materials and engineering. His work encompasses drafting, prosecution, opposition, dispute resolution and litigation – all stages of the patent life cycle. Matthew has a degree and PhD in materials science from the University of Oxford. His focus is on helping clients to navigate the opportunities and challenges of the Unified Patent Court.
Email: matthew.naylor@mewburn.com
.png)