At Evolve, we have built our own proprietary AI tools and agents, engineered for the unique requirements of pharmaceutical IP. Einstein is our proprietary AI agent for patent validity analysis. Einstein is a powerful tool that enables us to provide our clients with fast, cost-effective competitive intelligence and third-party patent review. Whether you need to assess a competitor’s portfolio, review a potentially blocking third-party patent, identify freedom to operate risks, or complete diligence on an asset, Einstein’s analysis quickly provides a clear strategic view. The core of Einstein is its Data module, which extracts and analyses the data in any biotech or pharma patent document, to efficiently provide rapid and commercially useful information.
When you need a fast, high-level view, not a full opinion
Companies and in-house IP teams frequently need a quick, comprehensive read on a third party’s IP, or on the competitive landscape, to inform a decision or satisfy investors. In many cases, this is not the moment to spend large sums on a full freedom-to-operate (FTO) clearance or an extensive invalidity opinion. What is needed is a strategic view, fast: which third-party rights matter, how strong they really are, and where the risks and opportunities sit.
The problem for biotech and pharma companies is that, traditionally, even a high-level view has required labour-intensive, expensive analysis. Einstein changes that. It enables us to perform a deep, data-level analysis of third-party patents very quickly and cost-effectively, so we can give you a comprehensive high-level view when you need it, without the cost and time associated with a full FTO opinion.
What Einstein does
- Competitive intelligence: Einstein lets us map an entire competitor portfolio or technology landscape at speed, pulling the key experimental data from across many patents at once. This shows where a competitor’s real strength lies, identifying any field-blocking IP and new IP opportunities. We can also identify a competitor’s lead molecules directly from their patent filings to sharpen the strategic picture.
- Third-party IP review: Whether a competitor patent, in-license target, IP portfolio, or contentious matter, Einstein assists in a rapid in-depth review to support freedom-to-operate, due diligence, invalidity and infringement analysis. Einstein provides fast, evidence-based analysis to find patent vulnerabilities, map data to enforceable claim scope and help guide the decisions in front of you.
Importantly, Einstein for Data does not replace attorney review. Einstein is instead a powerful tool that provides the first detailed analysis and facilitates a deeper dive by the attorney into the complexities of the data and claim scope.
Built by pharma IP experts
Einstein is built by our own pharma patent attorney experts. As a purpose-built AI tool for pharma IP, it follows the science and knows the law. Einstein reads figures, data and complex experimental reports to test whether the science genuinely supports the claims, and draws on a knowledge bank curated by our experienced patent attorneys and case-law specialists on the evidence requirements for pharma and biotech patents. Einstein does the heavy lifting of data extraction, allowing us to focus on the strategic and commercially relevant questions. The result is the strategic review our clients actually need. With Einstein we can efficiently provide a clear, data-grounded answer to the business questions: is there any field-blocking IP, is this patent a genuine threat (or opportunity), is it vulnerable to challenge?
Each of the three examples below shows the output from Einstein on real-world (anonymised) examples. For each case, the output was verified as being objectively correct. It shows how Einstein is a powerful tool helping us efficiently dive deeply into the data.
These analyses can form the basis of an expert review into the data, validity and scope of the patent, assisting rapid analysis of competitive landscapes and IP portfolios.
A complete AI Ecosystem for pharma IP
Einstein for Data is just one part of a wider suite of bespoke AI agents Evolve has built for pharmaceutical IP, covering drafting, prosecution, licensing, portfolio visualisation, and disclosure review. As pharma IP specialists, we have thought deeply about what is actually useful day-to-day for pharma in-house teams. We are building systems that solve the problems our clients have and enable us to focus on providing strategic advice.
As your fractional head of IP, or an extension of your existing in-house team, we use Einstein to deliver competitive intelligence, freedom-to-operate, and deal-readiness work to a high calibre, fast, and at a fraction of the cost and time of conventional review.
Example 1
Einstein’s anonymised analysis of a patent document. Einstein mapped each embodiment to its claimed technical effect, the supporting experimental evidence and the claims it underpinned, and scored the strength of that evidence. The analysis shows that the lead siRNA LNP anti-TX Construct 103 was the strongest embodiment, backed by in vivo data showing targeted cell depletion and tumour control and supporting the broadest claims. A middle tier of constructs and formulations rested on in vitro data alone, whilst the alternative payloads were merely exemplified, with no supporting data and underpinning no claims at all.
Einstein’s siRNA LNP IP data analysis:
| Embodiment | Technical Contribution | Supporting Data | Data Score | Claims |
| Anti-TX siRNA LNP (SEQ ID NO: 402) Short Interfering RNA (siRNA) targeting TX, encapsulated in an LNP. Construct 103. | Depletion of TX-expressing target cells in an in vivo disease model. | 🐁In Vivo. “The siRNA targeting TX mediated specific knockdown and depletion of target cells in the humanized mice. Reduction in TX-expressing cells was maintained across the doses within the peripheral blood, spleen, bone marrow and lymph nodes.” (Detailed Description, Page 85, Para [0341]). | 1 🏅 | Claims 1-35 |
| Reduction of disease-associated biomarkers. | 🐁In Vivo. “Disease biomarker levels were collected from the plasma sample at week 2, week 8 and week 12 after disease induction and quantified by a commercially available enzyme-linked immunosorbent assay (ELISA)… Resulting biomarker levels in the blood are shown in Figs. 5A-5D.” (Detailed Description, Page 85, Para [0342]). | |||
| Anti-TX siRNA LNP (General) Short Interfering RNA targeting TX, encapsulated in an LNP (general formulation). | Depletion of target cells in humanized mice. | 🐁In Vivo. “The siRNA targeting TX mediated specific depletion of target cells in humanized mice. Reduction in TX-expressing cells was maintained upon 5 doses within the peripheral blood and spleen.” (Detailed Description, Page 82, Para [0326]). | 2 | Claims 1-35 |
| Control of tumor burden via knockdown of TX. | 🐁In Vivo. “Tumor clearance was observed in this mouse model. Mice treated with anti-TX siRNA encapsulated in LNP exhibited tumor control until the study endpoint at Day 24.” (Detailed Description, Page 83, Para [0330]). | |||
| Anti-TY siRNA LNP Short Interfering RNA targeting alternative target TY, encapsulated in an LNP. | Confirmation of TY expression in target tissues. | 🐁In Vivo. “TY was detected on the surface of target cells in the bone marrow and spleen at each timepoint, with peak expression observed at 72 hours post LNP-siRNA treatment.” (Detailed Description, Page 83, Para [0331]). | 3 | ❌ None |
| Anti-TX siRNA Constructs 1-85 Constructs testing combinations of various guide sequences and chemical modifications. | TX knockdown and siRNA stability in primary human T cells. | 🔬In Vitro. “Starting at 24 hours post electroporation up to 120 hours, TX expression (T cell MFI), i.e., the level of expression per cell over time, was evaluated via flow cytometry to quantify knockdown…” (Detailed Description, Page 75, Para [0215]). | 4 | Claims 1, 3-4, 8-20, 28-35 |
| Target-specific cytotoxicity against cell lines and cytokine production. | 🔬In Vitro. “Cell killing was observed in T cells transfected with anti-TX siRNA-LNP using two donor samples.” (Detailed Description, Page 83, Para [0327]). “At 24-hour and 48-hour timepoints, supernatants were collected and cytokine amounts were quantified…” (Detailed Description, Page 78, Para [0237]). | |||
| Anti-TZ siRNA Constructs Constructs targeting alternative target TZ with various siRNA designs. | Knockdown of TZ in human T cells. | 🔬In Vitro. “The donor T cells were electroporated with the siRNAs at either 10 ng, 30ng, or 100 ng per 0.1×10^6 T cell … TZ expression was assessed using fluorescence activated cell sorting (FACS).” (Detailed Description, Page 79, Para [0252]). | 4 | Claims 1-2, 4, 28-35 |
| Target-specific cytotoxicity against TZ-expressing cells. | 🔬In Vitro. “24 hours later, the T cells were cocultured with TZ-expressing cell lines at E:T = 1:1 and 2:1 for target specific cytotoxicity for a span of three days via total accumulation of cell death… Anti-TZ siRNA-treated cells showed target specific killing.” (Detailed Description, Page 78, Para [0249]). | |||
| Lipid Nanoparticle Formulations 1-15 LNP mixtures of varying molar ratios of ionizable lipids, helper lipids, structural lipids, and PEG-lipids. | Encapsulation of siRNA into stable lipid nanoparticles. | 🔬In Vitro. “Final concentration=1.35 mg/mL siRNA (encapsulated). Zave=75.9 nm (Dv(50)=57.3 nm; Dv(90)=92.1 nm).” (Detailed Description, Page 52, Table 4c). | 5 | Claims 5-7, 21-27 |
| Exemplary-Only Therapeutics Alternative payloads including siRNAs against unrelated antigens, transcription factor mRNAs, enzymes and cytokines. | Expression or knockdown of corresponding therapeutic targets. | ❌ No data. None. | 6 | ❌ None |
Example 2
Einstein’s anonymised analysis of a patent for “AXL404, a humanised anti-CDX22a antibody and its targeted LNPs”. Einstein found that the lead AXL404 whole antibody and tLNP was the strongest embodiment, backed by both in vivo and in vitro data and underpinning the broadest claims. A tier of engineered F(ab’) analogues and antibody variants followed, resting largely on in vitro data, with several flagged for reduced developability, thermostability or binding liabilities. The untested AXL variants, by contrast, had no supporting data at all, yet still reached into the claims.
Einstein’s tLNP IP data analysis:
| Embodiment | Technical Contribution | Supporting Data | Data Score | Claims |
| AXL404 Whole Antibody & tLNPs (Humanized IgG1-FcAla anti-CDX22a, SEQ ID NO: 104/105) | Binds human and non-human primate CDX22a with high affinity, mediating efficient cross-species targeting and transfection in vivo and in vitro. | 🐁In Vivo & 🔬In Vitro “Every evaluated CDX22a-targeting construct demonstrated adequate binding kinetics, with KD measurements falling between 2.15 nM and 11.40 nM” (Examples, Page 112, Para [00315]). “The intact AXL404 immunoglobulin displayed comparable half-maximal effective concentrations (EC50) across T lymphocytes derived from humans and macaques” (Examples, Page 116, Para [00325]). “Across the entire range of evaluated targeting-ligand-to-payload ratios, successful reporter gene translation was observed specifically within the CDX22+ T cell subset, while CD4+ populations remained untransfected” (Examples, Page 116, Para [00326]). | 1 🏅 | 1, 4-9, 12-15, 30-36, 45-50, 52-60 |
| Exhibits superior biophysical developability, lacking self-aggregation, deamidation (unstressed), and polyreactivity. | 🔬In Vitro “Among the half-dozen engineered constructs evaluated in full-length format, only variants AXL404, AXL406, AXL410, and AXL411 successfully satisfied the stability criteria” (Examples, Page 118, Para [00340]). “While minor off-target interactions were detected for AXL403 at elevated concentrations (200 µg/mL), the AXL404 construct remained highly specific under identical conditions” (Examples, Page 119, Para [00345]). | 1 🏅 | 1, 4-9, 12-15, 30-36, 45-50, 52-60 | |
| Disulfide-Engineered F(ab’) Analogues (.12, .18, .22, .24) (e.g., AXL404.18, AXL404.22) | Features a relocated interchain disulfide bond allowing site-specific LNP conjugation via a free thiol without structural disruption, yielding potent in vivo transfection comparable to whole antibody. | 🐁In Vivo & 🔬In Vitro “Evaluation in humanized mouse models confirmed that targeted lipid carriers functionalized with the modified F(ab’) domains achieved in vivo delivery rates matching those of full-length IgG controls” (Examples, Page 125, Para [00371]). “…whereas architectural variants .22 and .24 demonstrated robust and reproducible performance profiles” (Examples, Page 126, Para [00375]). | 1 🏅 | 16-29, 37-44, 51-60 |
| ZKX1 tLNP Targeting Moiety (Humanized anti-CDX22, SEQ ID NO: 311/312) | Targets the membrane-proximal CX-Alpha structural epitope, mediating highly efficient payload expression in vivo. | 🐁In Vivo “Systemic administration in human-engrafted murine models revealed that nanoparticle formulations decorated with either ZKX1 or AXL404 achieved practically equivalent levels of target cell modification” (Examples, Page 128, Para [00382]). | 1 🏅 | 38, 39, 43, 51-60 |
| Liability-Engineered F(ab’) Variants (e.g., AXL711, AXL712 with N42Q/A and E22T mutations) | Eliminates specific deamidation and isomerization liabilities in CDRs, significantly improving binding affinity while maintaining uniform transfection efficiency. | 🔬In Vitro “Constructs harboring the dual amino acid substitutions exhibited binding strengths that were two to four times greater than the parental AXL404 baseline” (Examples, Page 121, Para [00360]). “…ex vivo assays comparing nanoparticles directed by liability-optimized domains (AXL711, AXL712) against those using the base AXL404 sequence showed no significant disparity in either the frequency of modified cells or the density of surface receptor expression” (Examples, Page 126, Para [00375]). | 4 | 1, 2-5, 8-10, 16-29, 37-44, 51-60 |
| AXL406 & AXL410 Whole Antibodies (Humanized IgG1-FcAla variants) | Provides specific CDX22 targeting with acceptable affinity, high thermostability (Tm > 65°C), and negligible polyreactivity. | 🔬In Vitro “Among the half-dozen engineered constructs evaluated in full-length format, only variants AXL404, AXL406, AXL410, and AXL411 successfully satisfied the stability criteria” (Examples, Page 118, Para [00340]). “…variants AXL404, AXL406, and AXL410 demonstrated strict target exclusivity, successfully maintaining the highly specific binding profile of their predecessor” (Examples, Page 119, Para [00348]). | 5 | 1, 4-9, 12-15, 30-36 |
| AXL505 Whole Antibody (N42D deamidation mimic) | Models 100% deamidation at VH N42, retaining substantial but statistically reduced delivery capability. | 🔬In Vitro ⚠️ “…cells treated with the engineered N42D substitute still showed notable reporter activity, though it was measurably lower (a 15-20% decrease) than the reference AXL404 formulation” (Examples, Page 121, Para [00359]). | 6 | 1, 3, 5, 8-10, 12-15 |
| AXL403 & AXL405 Whole Antibodies (Humanized IgG1-FcAla variants) | Achieves in vitro binding and transfection, but exhibits deficient thermostability and marginal polyreactivity. | 🔬In Vitro ⚠️ “While several candidates cleared the thermal stability benchmark, the AXL403 and AXL405 full-length immunoglobulins failed to achieve the minimum required melting temperature” (Examples, Page 118, Para [00340]). ⚠️ “Elevated concentrations (200 µg/mL) of the AXL403 preparation revealed minor promiscuous binding tendencies” (Examples, Page 119, Para [00345]). | 7 | 1, 4-9, 12-15 |
| Disulfide-Engineered F(ab’) Design .21 (e.g., AXL404.21, AXL711.21) | Alternative C-terminal truncation format (P221) intended for LNP conjugation. | 🔬In Vitro ⚠️ “…formulations utilizing the .18 and .21 framework architectures occasionally yielded sub-optimal levels of translated payload” (Examples, Page 126, Para [00375]). | 8 | 16, 22, 25, 30, 37-44, 51-60 |
| BPT-44.1 tLNP Targeting Moiety (Anti-CDX22 antibody) | Competes for the favorable CX-Alpha structural epitope, conferring hypothetical efficient LNP targeting analogous to ZKX1 and AXL404. | 🔬In Vitro “The accumulated evidence implies that the BPT-44.1 clone, or a derived variable fragment, represents a highly capable alternative for directing therapeutic carriers to the CDX22 receptor” (Examples, Page 128, Para [00382]). | 9 | 38, 39, 44, 51-60 |
| Untested AXL Variants (e.g., AXL407-409, AXL412-421, AXL501-502) | Alternative framework-grafted humanized heavy and light chains for binding CDX22. | ❌ No data. | 10 | 1, 4-9, 12-15 |
Example 3
Einstein’s anonymised analysis of a patent for “a recombinant Foamy Virus (FV) vector targeting MUC16”. Here a single lead embodiment carried the patent, the FV-based anti-MUC16 vector was robustly supported by in vivo non-human primate data demonstrating efficacy, safety and successful NK cell transduction, and underpinned the broadest claims. Every other embodiment, the alternative viral envelope variants and the alternative solid-tumour antigen receptors, was entirely hypothetical, with no supporting data, despite reaching across much of the same claim scope, exposing a clear gap between what the patent claimed and what its data could support.
Einstein’s Foamy Virus vector data analysis:
| Embodiment | Technical Contribution | Supporting Data | Data Score | Claims |
| Recombinant Foamy Virus (FV) based vector targeting MUC16 Viral vector encoding an EF2-alpha promoter operably linked to an anti-MUC16 CAR (SEQ ID NOs: 405, 406). Viral envelope comprises a mutated Lyssavirus-G protein (S52P and D411G), a primary anti-CD28 non-viral membrane bound tropism polypeptide (KL99-4), and a secondary OX40L non-viral membrane bound tropism polypeptide. | Induces NK cell expansion and durable MUC16+ cell aplasia in a non-human primate model. | 🐁In Vivo Experiments in Papio anubis dosed at 2.5×10¹¹ vg/kg and 2.5×10¹⁰ vg/kg demonstrate efficacy. “NK cells transduced with the FV derived vector encoding the anti-MUC16 CAR potently reduced target cells and mediated durable MUC16+ cell aplasia.” (Examples, Page 42; Figure 2, Figure 4). | 1 🏅 | Claims 1-12, 15-20, 30-55 |
| Enables safe in vivo administration without neurotoxicity or cytokine release syndrome (CRS). | 🐁In Vivo Levels of cytokines/chemokines evaluated post-infusion. “Figure 5 shows that the in vivo FV derived vector was safe at 2.5×10¹¹ vg/kg and did not increase cytokine/chemokines to levels associated with neurotoxicity in P. anubis.” (Examples, Page 45; Figure 5). | |||
| Facilitates successful in vivo transduction of target NK cells. | 🐁In Vivo Vector copy number evaluated post-infusion. “Figure 4, right panel, shows the vector copy number (VCN) in NK cells transduced with the FV derived vector encoding the anti-MUC16 CAR.” (Examples, Page 46; Figure 4). | |||
| Lyssavirus-G Envelope Variants Recombinant viral vector comprising a Lyssavirus-G envelope protein having mutations at L12, S52, W250, and/or D411, or variants corresponding to SEQ ID NOs: 500-588. | Mediates fusion of the viral particle with the target host cell without binding the cognate receptor (TfR1). | ❌ No data | 2 | Claims 1-10, 15-20, 30-55 |
| Mokola-G Envelope Variants Recombinant viral vector comprising a Mokola-G envelope protein having mutations at S52 and/or D411, or variants corresponding to SEQ ID NOs: 600-620. | Mediates fusion of the viral particle with the target host cell without binding the cognate receptor. | ❌ No data | 2 | Claims 1-10, 21-25, 30-55 |
| Paramyxovirus Envelope Variants Recombinant viral vector comprising Paramyxovirus envelope glycoproteins (PaV-F and PaV-H), specifically PaV-H with mutations at W502, K550, T560, and Y561 (e.g., W502A, K550A, T560I, Y561A). | Mediates fusion of the viral particle with the target host cell without binding the cognate receptor. | ❌ No data | 2 | Claims 1-10, 26-28, 30-55 |
| Pneumovirus Envelope Variants Recombinant viral vector comprising Pneumovirus envelope glycoproteins (PnV-F and PnV-G), specifically PnV-G with mutations at D601, Y605, N620, and D625 (e.g., D601A, Y605A, N620A, D625A). | Mediates fusion of the viral particle with the target host cell without binding the cognate receptor. | ❌ No data | 2 | Claims 1-10, 29, 30-55 |
| Alternative Solid Tumor Engineered Antigen Receptors Recombinant viral vector encoding an engineered antigen receptor (CAR, CCR, TCR, DARIC, etc.) targeting alternative solid tumor antigens (e.g., Mesothelin, PSMA, CEA, HER2, EGFRvIII, Claudin-18.2, GD2). | Redirects immune effector cells to target alternative cell surface antigens to reduce cells of the solid tumor lineage. | ❌ No data | 2 | Claims 1-10, 13, 30-55 |
The Experts
Rose Hughes, MA (Cantab), PhD, EPA
Rose is a biotech and pharmaceutical patent specialist with over a decade of experience in intellectual property. At Evolve, Rose leverages our unique fractional in-house model to provide clients with deep patent law expertise combined with the strategic commercial oversight of senior in-house counsel. Rose’s client’s include late-stage cell therapy and clinical stage biotech companies.
With a PhD in Immunology from UCL, Rose applies her technical background to complex innovations in biologics, cell and gene therapies, and the rapidly emerging field of AI-assisted drug development. Previously, Rose held the role of Director. Patents at AstraZeneca, where she was responsible for global IP portfolios and IP strategy at every stage of the pharmaceutical pipeline, from platform development and on-market commercialization to SPCs and patent term extensions.
A recognized thought leader in the field, Rose has been a regular contributor to IPKat since 2018, offering practical insights into European patent law developments. She is Editor of the CIPA Guide to the Patents Act (10th Edition), a frequent speaker on the epi podcast, a guest lecturer for the Brunel University IP law Postgrad Certificate, and a contributing author to published books A User’s Guide to Intellectual Property in Life Sciences (2021) and Developments and Directions in Intellectual Property Law (2023).
Steven Gurney, EPA, CPA, MIPLaw
Steven is focused on maximising the potential of his pharmaceutical and life science clients. As the founder of Evolve Intellectual Property in 2021, he brings a passion for developing robust IP strategies and providing expert counsel.
A highly skilled life science patent attorney with over 20 years of combined in-house and private practice experience, Steven is qualified as a European, UK, and Australian patent attorney. His recognition in legal directories includes LEGAL 500 (2024, 2025), IP STARS (2024, 2025), and IAM Patent 1000 (2024).
Steven’s expertise spans pharmaceuticals, drug development and IP, drafting, portfolio management and prosecution, lifecycle management, licensing (in- and out-), due diligence and FTO, contentious work (including litigation and opposition), and regulatory exclusivities like PTE/SPC.
