AI Drug Repurposing Patents

AI drug repurposing patents cover method-of-use/new-indication innovations; AI-prediction/knowledge-graph innovations; formulation/combination innovations; and biomarker/patient-selection and platform/data innovations — with IP held by AI-repurposing companies and pharma (in a field finding new uses for existing drugs). WHY AI DRUG REPURPOSING: it uses AI to find NEW uses for EXISTING drugs — taking a drug already APPROVED (or shelved after a failed trial) for one disease and discovering it ALSO treats a DIFFERENT disease; repurposing is attractive because the drug's SAFETY is already established (it's been in humans), so development is FASTER, CHEAPER, and far LESS RISKY than developing a brand-new molecule; AI accelerates the search by mining vast biomedical data — KNOWLEDGE GRAPHS that link drugs, targets, genes, diseases, pathways, and the scientific literature — and machine-learning models that PREDICT 'this existing drug should work on that disease' by surfacing hidden connections humans would miss. THE CENTRAL IP PROBLEM that makes repurposing uniquely tricky: you usually CANNOT get COMPOSITION-OF-MATTER protection (the molecule is old, known, and often off-patent), so the patentable IP is mostly METHOD-OF-USE (a new indication) and method-of-treatment claims — which are NARROWER, easier to design/prescribe around (off-label use), and (for the AI predictions and any natural correlation) face §101 — making the exclusivity strategy fundamentally different and harder than for novel-molecule drug discovery. MAJOR HOLDERS/PLAYERS: HEALX, BENEVOLENTAI, RECURSION, EVERY CURE, plus pharma. Method-of-use/new-indication, AI prediction/knowledge graph, formulation/combination, biomarker/patient selection, and platform/data are the core repurposing patent domains — with the composition-of-matter gap and §101 the constraints, and method-of-use, formulations, biomarkers, and the platform the open whitespace.

Method-of-use/new-indication innovations; AI-prediction/knowledge-graph innovations; formulation/combination innovations; and §101-aware claiming represent core repurposing patent domains — and the new-indication method and the AI discovery engine are the foundational, high-value (but constrained) capabilities. METHOD-OF-USE / NEW-INDICATION PATENTS: the CORE patentable IP for repurposing — a NEW medical USE for a KNOWN drug, claimed as a METHOD OF TREATING disease Y by administering drug X (since composition-of-matter on the old molecule is usually unavailable); method-of-use/new-indication claims are the central, high-value (but NARROWER) IP (the new indication is the invention — but method-of-use patents are inherently narrower than composition-of-matter, can be circumvented by off-label prescribing/generics, and need careful claiming, so they're real but limited protection — the defining repurposing IP challenge). AI-PREDICTION / KNOWLEDGE-GRAPH PATENTS: the discovery ENGINE — KNOWLEDGE GRAPHS integrating drugs/targets/genes/diseases/literature, and ML MODELS that predict new drug-disease links (target prediction, link prediction, similarity-based repurposing); AI-prediction/knowledge-graph methods are high-value IP BUT §101-SENSITIVE (claim a specific technical system/architecture, not the abstract idea of 'predict a drug for a disease') — and much of the value here is in the PROPRIETARY knowledge graph/data and models, often better protected as trade-secret/data than patents. FORMULATION / COMBINATION PATENTS: a NEW FORMULATION, dose, delivery route, or drug COMBINATION for the new use — adding patentable angles and exclusivity AROUND the old molecule (a reformulation or combination can be a fresh composition/method); formulation/combination methods are high-value, distinctive IP (formulations/combinations are a key way to build defensible IP and exclusivity for a repurposed drug beyond a narrow method-of-use). §101-AWARE CLAIMING: claim the new-use treatment METHOD, a specific technical AI system, and concrete formulations — not abstract prediction or a natural correlation; §101-aware claiming matters. Method-of-use/new-indication, AI prediction/knowledge graph, formulation/combination, and §101-aware claiming are the highest-value core IP because a new indication, a technical discovery engine, and reformulations/combinations are exactly what is patentable for an old drug in a new use.

Biomarker/patient-selection innovations; platform/data innovations; exclusivity-strategy innovations; and combination innovations represent additional repurposing patent domains — and patient selection, the data/platform moat, and regulatory exclusivity are where repurposing builds defensibility beyond narrow patents. BIOMARKER / PATIENT-SELECTION PATENTS: a BIOMARKER or patient-selection method that defines WHICH patients the repurposed drug works for — making the new use more novel, specific, and protectable (a method of treating biomarker-positive patients with the drug); biomarker/patient-selection methods are high-value IP BUT §101-AWARE (a natural correlation is unpatentable — claim a treatment method using the biomarker, not the correlation). PLATFORM / DATA PATENTS: the repurposing PLATFORM and the PROPRIETARY knowledge graph/data/models — the discovery asset; platform/data value is high but often best protected as TRADE-SECRET/data (a comprehensive, curated biomedical knowledge graph and trained models are a major, hard-to-replicate asset — frequently the real moat, and data/models aren't directly patentable). EXCLUSIVITY-STRATEGY PATENTS: building exclusivity around an old drug via REGULATORY mechanisms — new-indication regulatory EXCLUSIVITY, orphan-drug designation (for rare diseases — a strong fit for repurposing), and 505(b)(2)/reformulation pathways — which can matter MORE than patents for a known molecule; exclusivity-strategy methods/approaches are high-value (regulatory exclusivity, esp. orphan designation, is often the most important protection for a repurposed drug — patents alone are weak, so the strategy combines method-of-use patents + regulatory exclusivity + formulation). COMBINATION PATENTS: novel drug COMBINATIONS for the new use (an old drug + another agent can be a patentable, harder-to-circumvent combination); combination methods are high-value IP. Biomarker/patient selection, platform/data, exclusivity strategy, and combinations are the highest-value application IP because patient selection, a proprietary data moat, regulatory exclusivity, and combinations are exactly what make a repurposed drug defensible when composition-of-matter is unavailable.

AI drug repurposing startup IP strategy must navigate the composition-of-matter gap (the #1 issue — the molecule is old/known/off-patent, so you usually CAN'T get the strong composition-of-matter protection that anchors normal drug IP; the entire strategy must work around this), the method-of-use limitation (method-of-use/new-indication patents are the core IP but are NARROWER than composition-of-matter and can be undercut by off-label prescribing and generics — real but limited), the regulatory-exclusivity importance (new-indication regulatory exclusivity and especially ORPHAN-DRUG designation often matter MORE than patents for a repurposed drug — combine patents with regulatory exclusivity), the §101 constraints (the AI predictions and any biomarker correlation face §101 — claim technical systems and treatment methods), the data/platform-as-the-real-moat reality (the proprietary knowledge graph, data, and models are often the biggest, most-durable advantage — protect as trade-secret/data, frequently more valuable than the narrow patents), the formulation/combination angle (reformulations and combinations build additional, harder-to-circumvent exclusivity around the old molecule), the biomarker/patient-selection angle (defining who the drug works for adds novelty/protectability), the business-model reality (repurposing's value is faster/cheaper development and the platform's deal flow — but the weak IP can make monetization hard, so partnering/regulatory exclusivity matters), and a landscape where method-of-use, AI prediction, formulation/combination, biomarkers, and the platform/data are the durable assets; understand that composition-of-matter is unavailable, so the durable IP/exclusivity is in method-of-use claims + regulatory exclusivity (esp. orphan), formulations/combinations, biomarker patient-selection, and the proprietary platform/data (trade-secret) — with regulatory exclusivity, the data/platform, formulations, and combinations often the real moat (not patents alone), and that regulatory exclusivity, the data moat, method-of-use breadth, formulation/combination, and clinical proof matter as much as patents; identify whitespace in formulations, combinations, biomarkers, and the data platform. AI DRUG REPURPOSING STARTUP IP STRATEGY: METHOD-OF-USE + REGULATORY EXCLUSIVITY, FORMULATIONS/COMBINATIONS, BIOMARKER SELECTION, AND THE PROPRIETARY PLATFORM/DATA ARE THE IP: patent method-of-use/new-indication, formulations/combinations, and biomarker patient-selection — and lean heavily on regulatory exclusivity and a trade-secret data/platform; COMPOSITION-OF-MATTER IS USUALLY UNAVAILABLE — THE WHOLE STRATEGY WORKS AROUND IT: the molecule is old/known/off-patent, so you can't get the strong composition-of-matter protection that anchors normal drug IP; METHOD-OF-USE IS THE CORE IP BUT NARROW: new-indication/method-of-treatment claims are central but narrower than composition-of-matter and undercut by off-label/generics — real but limited; REGULATORY EXCLUSIVITY (ESP. ORPHAN) OFTEN MATTERS MORE THAN PATENTS: new-indication exclusivity and orphan-drug designation are frequently the most important protection — combine with method-of-use patents; THE DATA/PLATFORM IS OFTEN THE REAL MOAT — TRADE-SECRET IT: a proprietary, curated knowledge graph + trained models are the biggest, most-durable advantage — often more valuable than the narrow patents (and not directly patentable); §101 CONSTRAINS THE AI + BIOMARKERS: predictions and biomarker correlations face §101 — claim technical systems and treatment methods, not abstractions/correlations; FORMULATIONS/COMBINATIONS BUILD EXCLUSIVITY AROUND THE OLD MOLECULE: reformulations and combinations are harder-to-circumvent additional IP; BIOMARKER/PATIENT-SELECTION ADDS NOVELTY: defining who the drug works for makes the new use more protectable (§101-aware); BUSINESS MODEL — WEAK IP COMPLICATES MONETIZATION: faster/cheaper development and platform deal flow are the value, but weak exclusivity makes monetization hard — partnering and regulatory exclusivity matter; EXCLUSIVITY/DATA/METHOD-OF-USE/FORMULATION/CLINICAL MATTER AS MUCH AS PATENTS: regulatory exclusivity, the data moat, method-of-use breadth, formulation/combination, and clinical proof drive value; WHEN TO PATENT (OR USE EXCLUSIVITY/DATA): NOVEL USE/FORMULATION/COMBINATION/BIOMARKER WITH CLINICAL/MECHANISTIC SUPPORT: file method-of-use/formulation/combination (and pursue orphan/regulatory exclusivity + trade-secret the platform) once a repurposing candidate shows support (predicted + mechanistically/clinically validated new use + biomarker-defined population + formulation/combination novelty) — a validated new use, regulatory exclusivity, and the data platform are the critical repurposing IP metrics; KEY FTO CHECKLIST: Healx/BenevolentAI/Recursion/Every Cure + pharma; composition-of-matter gap (old/off-patent molecule); method-of-use/new-indication (method of treating disease Y with drug X — narrow); AI prediction/knowledge graph (drug-disease link prediction — §101, trade-secret data/models); formulation/combination (new formulation/dose/route/combination); biomarker/patient selection (treatment method using biomarker — §101-aware); platform/data (knowledge graph/models — trade-secret); regulatory exclusivity (new-indication/orphan-drug/505(b)(2)); off-label/generic circumvention risk.