This week’s AI in Healthcare and Digital Health brief distills AI-driven developments in biopharma, digital health, clinical research, and policy across global and US markets, with concise takeaways.
In Today’s Newsletter
Dive deeper
🤝 Lilly expands XtalPi tie-up into AI bispecifics [1] [05 Nov 2025]
Context: Builds on 2023 small-molecule pact; Ailux platform uses structural modeling, generative design, developability analytics.
Key point: Lilly pays Ailux “double-digit millions” upfront/near-term, can nominate multiple target pairs and optionally license the platform, total potential $345M.
Implication: Partnerships/BD: Signals pipeline investment and modality expansion.
🧠 Roche bets $55M upfront on Manifold BBB shuttles [2] [03 Nov 2025]
Context: Manifold’s in vivo mDesign engine to discover BBB shuttles; Roche to carry preclinical/clinical; complements Roche Brainshuttle programs.
Key point: Deal includes up to $2B milestones plus royalties; Manifold retains use outside licensed targets.
Implication: Partnerships/BD: Signals pipeline investment and modality expansion.
💊 Omada Health to prescribe GLP-1s in 2026, scales companion care [3] [07 Nov 2025]
Context: 831k members, +53% YoY; GLP-1 track supports 100k+ members; positive adjusted EBITDA in Q3.
Key point: Nationwide prescribing next year, pairing licensed providers with behavior change programs amid expected price shifts.
Implication: DTC/telehealth: Could streamline initiation and adherence via remote prescribing and logistics.
🧬 Flatiron brings LLM-RWD quality framework to ESMO AI [4] [06 Nov 2025]
Context: Five studies on LLM extraction across systems/countries, GDPR-pseudonymized networks, and aNSCLC digital twins.
Key point: VALID framework and pan-tumor extraction methods aim to make LLM-derived oncology RWD trustworthy at scale.
Implication: Observational/RWE: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🧪 Insilico pivots into cardiometabolic longevity programs [5] [07 Nov 2025]
https://longevity.technology/news/insilico-expands-ai-reach-into-metabolic-health/
Context: 8 oral small molecules across GLP-1RA, NLRP3, NR3C1, GIPR, Amylin, APJ, Lp(a); positioned for low-dose combos.
Key point: Pharma.AI-designed candidates aim for improved safety/PK; two oral GLP-1RAs near preclinical candidate stage.
Implication: Partnerships/BD: Signals pipeline investment and modality expansion.
🧫 Evaxion unveils AI-designed AML vaccine EVX-04 [6] [03 Nov 2025]
Context: Targets non-conventional ERV antigens; preclinical; ASH presentation Dec 6.
Key point: Off-the-shelf concept intended for rapid post-diagnosis use; company cites induced T-cell responses in preclinical data.
Implication: Partnerships/BD: Signals pipeline investment and modality expansion.
🧑⚕️ Hims & Hers scales AI-personalized, subscription care model [7] [07 Nov 2025]
Context: Expands globally and into Canada aligned with expected generic semaglutide; completed $870M convert notes; ZAVA acquisition extends EU reach.
Key point: Vertically integrated fulfillment plus AI personalization to support lower-cost GLP-1 programs at scale.
Implication: DTC/telehealth: Could streamline initiation and adherence via remote prescribing and logistics.
🤖 Harbour BioMed–Evinova China AI collaboration [8] [07 Nov 2025]
https://lelezard.com/en/news-22007828.html
Context: Aims to apply AI/digital tools to innovative biologics development, leveraging Harbour Mice and Evinova platforms.
Key point: Partners plan an open ecosystem for AI-driven discovery and clinical efficiency, rooted in China with global focus.
Implication: Partnerships/BD: Signals pipeline investment and modality expansion.
🫀 Cardiosense validates AI congestion tool vs physical exam [9] [06 Nov 2025]
Context: Secondary analysis from SEISMIC-HF I; moderated poster at AHA 2025 in New Orleans.
Key point: Early analysis indicates AI noninvasive PCWP classification could exceed standard physical exam for congestion.
Implication: Observational/RWE: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
👁️ Monash launches retinal AI for systemic disease screening [10] [06 Nov 2025]
Context: With DHCRC and Optain Health, building multimodal foundational model using large retinal and linked health datasets.
Key point: Seeks noninvasive biomarkers for CVD and CKD risk stratification using oculomics.
Implication: Access programs: May expand screening, initiation, and follow-up at scale.
☢️ Lumonus raises A$25M for AI radiation oncology workflows [11] [07 Nov 2025]
Context: Backed by Aviron and Oncology Ventures; platform used across US, Australia, EU; 280k treatments supported, 75k plans automated.
Key point: Funds product expansion and informatics, plus US go-to-market and RWE partnerships with health systems.
Implication: Access programs: May expand screening, initiation, and follow-up at scale.
🍽️ Noom debuts diabetes program with AI glucose forecasting [12] [07 Nov 2025]
Context: Employer/plan offering with coaching and nutrition; AI predicts post-meal glucose without CGM.
Key point: Validation cited: ~1.0% A1C drop and −29 mg/dL glucose over 3 months in program cohort.
Implication: DTC/telehealth: Could streamline initiation and adherence via remote prescribing and logistics.
🇮🇳 PB Health acquires Fitterfly to scale AI chronic care [13] [06 Nov 2025]
Context: Part of PB Fintech’s integrated care plan with hospitals, insurers, and preventive services; $218M funding in May 2025.
Key point: Fitterfly’s validated programs anchor diabetes, obesity, heart health strategy; AI tools embedded across workflows.
Implication: Access programs: May expand screening, initiation, and follow-up at scale.
🧠 ‘Mind-captioning’ turns brain activity into text [14] [06 Nov 2025]
https://www.nature.com/articles/d41586-025-03624-1
Context: Science Advances study uses fMRI with AI to decode scene meaning signatures into sentences.
Key point: Descriptions align with what participants see or imagine, without relying solely on generative structure.
Implication: Observational/RWE: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🌐 Lung cancer patient virtual communities review [15] [07 Nov 2025]
https://journals.sagepub.com/doi/10.1177/20552076251395985
Context: Scoping review of 19 studies (2015–2024).
Key point: Apps/web communities reduced anxiety and improved adherence, but no survival benefit; misinformation/privacy risks noted.
Implication: Observational/RWE: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🧩 PLOS: General-purpose generative model for healthcare data [16] [05 Nov 2025]
https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001059
Context: Integrates multiple clinical studies to model >2000 attributes; API supports imputation/synthetic data.
Key point: Captures key univariate/bivariate properties, demonstrated in practical apps.
Implication: Observational/RWE: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🩻 3DINO: self-supervised 3D medical imaging foundation model [17] [07 Nov 2025]
https://www.nature.com/articles/s41746-025-02035-w
Context: Pretrained on ~100k multimodal 3D scans; ViT adapted with 3D objectives.
Key point: Outperforms SOTA on diverse downstream tasks, reduces labeled data needs 4–10×.
Implication: Partnerships/BD: Signals pipeline investment and modality expansion.
🎓 RAG teaching assistant in medical education [18] [04 Nov 2025]
https://www.nature.com/articles/s41746-025-02022-1
Context: Deployed across two cohorts; constrained to instructor-curated materials.
Key point: Usage spikes pre-exams; trust rises with citations, but scope limits frustrate some users.
Implication: Access programs: May expand screening, initiation, and follow-up at scale.
🧻 IBD-BOOST digital CBT RCT shows no primary benefit overall [19] [07 Nov 2025]
https://www.thelancet.com/journals/landig/article/PIIS2589-7500(25)00088-3/fulltext
Context: n=780 UK multicenter RCT; 6- and 12-month outcomes vs care as usual.
Key point: No significant difference on IBD-Q or global relief; some benefit in faecal incontinence and generic QoL; compliers did better.
Implication: Observational/RWE: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
📈 Viewpoint: Using CGM in prediabetes behavioral research [20] [03 Nov 2025]
https://www.thelancet.com/journals/landig/article/PIIS2589-7500(25)00086-X/fulltext
Context: Review of 19 studies and methods.
Key point: Calls for clearer sampling, feedback protocols, standardized metrics, and behavioral principles in CGM-based interventions.
Implication: Observational/RWE: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🧪 npj AI: Multimodal AI set to reshape oncology [21] [07 Nov 2025]
https://www.nature.com/articles/s44387-025-00044-4
Context: Perspective on MMAI across screening, diagnosis, trials, RWE, and economics.
Key point: Highlights TRIDENT and ABACO case studies, synthetic controls, and payer alignment potential.
Implication: Observational/RWE: Could inform practice and payer discussions; interpretation depends on study design and confounding control.
🧩 Google Research: “Nested Learning” for continual learning [22] [07 Nov 2025]
https://research.google/blog/introducing-nested-learning-a-new-ml-paradigm-for-continual-learning/
Context: Frames models as nested optimization levels; introduces “Hope” architecture and continuum memory.
Key point: Reports improved long-context memory and reduced catastrophic forgetting in tests.
Implication: Partnerships/BD: Signals pipeline investment and modality expansion.
📉 “Densing law” of LLMs: capability density doubling trend [23] [06 Nov 2025]
https://www.nature.com/articles/s42256-025-01137-0
Context: Proposes capability-per-parameter metric; analyzes 51 OSS models.
Key point: Finds maximum capability density doubles ~every 3.5 months, with implications for cost and deployment.
Implication: Partnerships/BD: Signals pipeline investment and modality expansion.
🏛️ Telehealth policy: Remote e-prescribing cliff Dec 31 [24] [05 Nov 2025]
Context: ATA/ATA Action warn expiration of controlled-substance tele-prescribing flexibilities amid shutdown.
Key point: Calls for DEA/HHS extension and Congressional action; cites OUD access risks.
Implication: Access programs: May expand screening, initiation, and follow-up at scale.
❤️ ACC partners with OpenEvidence on AI point-of-care guidance [25] [07 Nov 2025]
Context: Generative AI CDS to synthesize ACC guidelines, with expert work group and resource center.
Key point: Aims for real-time, transparent, evidence-grounded recommendations across specialties.
Implication: Access programs: May expand screening, initiation, and follow-up at scale.
Why it matters
-
AI discovery/engineering deals continue shifting from small molecules to biologics, with bispecifics and BBB shuttles front and center [1][2].
-
GLP-1 care is moving from access to optimization as prescribing integrates with behavior change at scale [3][7][12].
-
Trustworthy AI-extracted RWD and multimodal analytics are maturing, informing evidence generation and payer decisions [4][21].
-
Foundation models in imaging and data synthesis promise label-efficient development and broader deployment in clinics [16][17][23].
-
Policy timelines, especially US tele-prescribing flexibilities, could materially alter virtual care continuity by year-end [24].
🚀 Accelerate your success. Contact us now
📂 Explore our case studies. See examples of our work.
💡 Read our insights. Learn from our latest reports and analysis
🎬 Watch on YouTube. Subscribe and never miss a video.
🧰 See our full range of services. Discover how we can help you.
📚 Find your one-stop page for the full AI and Digital Health archive.
FAQ
What exactly did Lilly get in its Ailux deal?
Lilly secured access to Ailux’s bispecific engineering platform with an upfront/near-term “double-digit millions,” options to nominate multiple target pairs, and an internal-use license option; total potential value up to $345M [1].
How will Roche and Manifold split work on BBB shuttles?
Manifold leads early discovery using its in vivo mDesign engine and shuttle portfolio; Roche takes preclinical and clinical forward, with $55M upfront and up to $2B in milestones plus royalties [2].
When will Omada start prescribing GLP-1s and what’s different?
Omada says nationwide prescribing in 2026, embedding licensed providers into its existing GLP-1 companion program that combines devices, coaching, and AI, targeting long-term outcomes and cost management [3].
What is Flatiron’s VALID framework?
VALID is a methodology to evaluate accuracy and quality of LLM-extracted oncology real-world data across systems and countries, showcased in multiple ESMO AI abstracts [4].
What targets are in Insilico’s cardiometabolic slate?
Programs include oral GLP-1RAs, selective NLRP3, NR3C1, GIPR, Amylin, APJ, and Lp(a), with two oral GLP-1RAs nearing preclinical candidate stage, all discovered via Pharma.AI [5].
What policy deadline is the ATA warning about?
US remote prescribing flexibilities for controlled substances via telehealth are set to expire on 31 Dec 2025 without DEA/HHS/legislative action, potentially disrupting OUD and mental health care access [24].
Entities / Keywords
Eli Lilly; XtalPi/Ailux; bispecific antibodies; Roche; Manifold Bio; BBB shuttles; Omada Health; GLP-1; Flatiron Health; VALID framework; Insilico Medicine; cardiometabolic; Evaxion; ERV antigens; Hims & Hers; Evinova; Harbour BioMed; Cardiosense; PCWP; Monash; oculomics; Lumonus; radiation oncology; Noom; PB Fintech / PB Health; Fitterfly; mind-captioning; virtual communities; VHGM; 3DINO; RAG teaching assistant; IBD-BOOST; CGM methods; MMAI oncology; Nested Learning; Densing law; ATA/ATA Action; ACC; OpenEvidence.
References
-
https://longevity.technology/news/insilico-expands-ai-reach-into-metabolic-health/
-
https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001059
-
https://www.thelancet.com/journals/landig/article/PIIS2589-7500(25)00088-3/fulltext
-
https://www.thelancet.com/journals/landig/article/PIIS2589-7500(25)00086-X/fulltext
-
https://research.google/blog/introducing-nested-learning-a-new-ml-paradigm-for-continual-learning/