This report reads every AD patent, publication and funded project, and company profiles, together. Scoring each target on both the strength of its evidence and how widely it is adopted shows what no single database gives you: which targets are validated but overlooked, which are crowded, and where patent activity is running ahead of the science.
It is built on AcademicLabs' comprehensive, continuously updated data. Our AI reads every record against disease-specific questions and stays grounded in the science: no web searches, no facts from model memory, and every figure linked back to its source record.
For your disease, target, technology or any topic.
Every atopic dermatitis (AD) target named in 3,070 AI-analysed patents, publications and funded projects, scored on two axes: how far up the evidence ladder it has climbed (from hypothesis to approved drug) and how widely industry and academia have taken it up. Built only from AcademicLabs data.
The IL-4/IL-13 axis alone holds 330 records and 198 of the 357 clinical-rung publication records (55%). Only 14 of 63 scored targets combine human or clinical evidence with broad adoption.
Just 34% of AD-relevant records carry human evidence (patient tissue, genetics or clinical data). 32 targets sit in "early signals": low evidence and low uptake.
STAT3, IRAK4, Bifidobacterium, MRGPRX2, NLRP3, IL-5, IL-17 and STAT6 draw many organisations while their evidence stays below the clinical rung. MRGPRX2: 37 patents (30 classed very recent) against 2 human-evidence records. IRAK4: 39 patents and 22 industry organisations.
Company answers carry evidence the structured fields miss: for IRAK4, the company profiles report phase 1 data for the degrader KT-474, including IRAK4 normalised in lesional skin and biomarker reductions in AD patients. Reading the text moves IRAK4 closer to the clinical rung than the patent and publication fields alone suggest.
IL-4, JAK1/JAK2, TYK2, IL-6, Lactobacillus, S. hominis, S1PR, IL-13Rα1, IL-22 reach human or clinical rungs but attract few organisations. In this group the AI analysis points to microbiome (29% of records) and combination approaches (30%) more often than in any other quadrant.
For validated targets the top challenge is safety (38% of records). For early-signal targets it is delivery and skin penetration (37%). Biomarker-guided stratification is the most-cited opportunity overall (456 records).
Trial registries rarely name the target. Publications do: 357 records sit at a clinical rung, including 60 randomised or phase 2/3 studies, 87 real-world studies and 32 meta-analyses. For 162 of them we extracted the result sentence.
One AcademicLabs workbook of AI-analysed exports on atopic dermatitis: 1,025 patents, 1,188 biology publications, 585 modality/clinical publications and 272 funded projects judged AD-relevant (of 3,449 records), plus a named-programme sheet (62 programmes) and a sample of 423 company profiles with AI-written answers on AD status, target, stage and data. Every record was analysed by AcademicLabs AI against a fixed question set; all aggregation below is deterministic code.
The primary target of each record was normalised (synonyms merged, e.g. OX40/OX40L, IL-31/IL-31RA, IL-33/ST2, KLK5/7). Category labels (pathways, cell types, "skin barrier") are excluded. 108 molecular targets are named; 63 with at least 5 records are scored.
Each record gets one rung from its structured fields: 0 concept/review, 1 in vitro/mechanistic, 2 animal, 3 human disease evidence (patient tissue, serum, genetics, ex vivo), 4 early clinical, 5 late clinical, 6 approved/marketed (named programmes only). Patents use Evidence stage and Evidence type; projects use development stage, evidence type and human target evidence; biology papers use strongest evidence type, human patient-derived evidence and development stage; modality papers use development stage, target maturity, study model and strongest evidence. The ceiling is the highest rung reached by at least two records or by a named programme, so a single passing mention cannot lift a target.
Evidence score (0-100) = 45 × ceiling/6 + 25 × human-evidence records + 15 × clinical records + 15 × biology-paper strength. Adoption score (0-100) = 35 × organisations + 30 × industry organisations + 15 × named programmes + 10 × countries + 10 × patent momentum. Counts enter on a log scale, scaled to the largest target, so that one very large target (IL-4/IL-13) does not flatten the rest. Quadrants split both scores at 45.
Why these weights. For evidence, how far a target has climbed matters most (45%): a target with human or clinical data has cleared the step where most AD targets fail, the translation from mouse to patient. The number of human-evidence records (25%) rewards confirmation across independent studies rather than a single report; clinical records (15%) add weight for interventional data; paper strength (15%), the share of biology papers our AI graded strong or moderate, stops a large volume of weak papers from inflating the score. For adoption, breadth comes first: how many organisations work on the target (35%), and especially how many companies (30%), since industry commitment signals that someone is paying to develop it. Named programmes (15%) count actual development assets, countries (10%) show geographic spread, and the share of very recent patents (10%) shows momentum. Scores rank targets within this dataset; they are not absolute measures of validation.
Publication records at a clinical rung were scanned for study design (keyword rules), sample size, drug names (INN suffixes and code patterns), endpoints (EASI, IGA, NRS, SCORAD, DLQI, POEM, TEWL, biomarkers) and the abstract sentence with the most quantitative content. These rows are auto-extracted, not curated: check the linked record before quoting.
Besides structured fields, AcademicLabs AI wrote a short judgement for every patent, publication, funded project and company profile: the challenges and opportunities it sees for the target, the main limitation of the evidence, and how the target is positioned against approved benchmarks. We grouped these judgements into 12 challenge and 12 opportunity themes and counted, per target and per quadrant, the share of records that raise each theme. Quotes in the dossiers are taken from the records highest on the evidence ladder. Company profiles also add evidence the structured fields miss, such as early clinical data disclosed by developers.
By evidence score: IL-4/IL-13, JAK1, IL-4Rα, IL-13, IL-31/IL-31RA, OX40/OX40L. All reach late-clinical or approved rungs with many human-evidence records. IL-4/IL-13 leads with 240 human-evidence records.
IL-4, JAK1/JAK2, TYK2, IL-6, Lactobacillus, S. hominis, S1PR, IL-13Rα1, IL-22: evidence at human or clinical level, but adoption below the midpoint (few organisations, patents or programmes).
STAT3, IRAK4, Bifidobacterium, MRGPRX2, NLRP3, IL-5, IL-17, STAT6. They draw many organisations while their evidence stays below the clinical rung. MRGPRX2 has 37 patents and 2 human-evidence records.
34% of the 3,070 AD-relevant records carry human evidence. 32 of 63 scored targets are still early signals.
From publications: 357 records sit at a clinical rung, 60 of them randomised or phase 2/3; 162 contain an extractable quantitative result.
Safety dominates for validated targets (38% of records); delivery and skin penetration for early-signal targets (37%). Biomarker-guided stratification is the most-cited opportunity overall.
Evidence (0-100): ladder ceiling 45%, human-evidence records 25%, clinical records 15%, paper strength 15%. Adoption (0-100): organisations 35%, industry organisations 30%, named programmes 15%, countries 10%, patent momentum 10%. Quadrants split at 45.
Prepared by Arne Smolders, Founder & CEO of AcademicLabs. Published 8 October 2026. Built with AcademicLabs from 3,070 AI-analysed records (patents, publications, funded projects) and a sample of 423 company profiles. Figures describe what the records state and are not medical or investment advice.
Coverage. Not included: clinical-trial registry records (they rarely name the target), Chinese-language patents and pipeline charts on company websites. The company profiles are a sample, so industry adoption from patents is the stronger signal.
Please cite and link this page:
Smolders A. Atopic Dermatitis Targets 2026: Evidence vs Adoption. AcademicLabs; 2026. https://academiclabs.com/research-reports/atopic-dermatitis-targets-evidence-adoptionThe data and figures are shared under CC BY 4.0: free to reuse with credit to AcademicLabs.