Federated network of 30 million patients powers first-series of 15 foundational studies from Metabolism Agentic Intelligence Atlas (MAIA) toward novel diagnostics, therapeutics, and surgical interventions.
CAMBRIDGE, Mass., June 9, 2026 /PRNewswire/ -- nference today announced a series of publications from the Metabolic Agentic Intelligence Atlas initiative, or MAIA, a federated AI program that enables real-time synthesis of biomedical knowledge from the de-identified care patterns of over 30 million patients. The studies provide collectively, for the first-time, a high-resolution precision map of the impact of GLP-1, GIP, and other incretin therapies on the whole human body beyond weight-loss and glucose measurements that have been the primary endpoints of clinical studies thus far. Across these studies, the nference MAIA team leveraged physician-validated Large Language Models (LLMs) and other AI technologies to derive deep insights into longitudinal clinical notes, medication exposure and adherence patterns, serial laboratory measurements, body weight and body-composition trajectories, radiology and pathology imaging, and electrophysiology waveforms to quantify healthy longevity outcomes.
"GLP-1 medicines are no longer just obesity drugs. They are therapeutic lenses into healthy longevity and evolution of diverse disease pathologies at a resolution that was unimaginable until recently," said Dr. Venky Soundararajan, PhD, Founder & Chief Scientific Officer of nference. "The consistent signal from the first set of MAIA studies is that next-generation clinical trials and decision-support systems should be designed with whole-body spatial intelligence."
Across the portfolio, MAIA is intended to go beyond what gold-standard randomized clinical trials (RCTs) have contributed, addressing a central blind spot in modern cardiometabolism: the field has measured pounds lost far more precisely than organs protected, disease trajectories redirected, or durability retained after therapy stops. MAIA turns routine care into a longitudinal discovery engine, connecting dose, adherence, discontinuation, tissue remodeling, infection resilience, immune modulation, neuropsychiatric outcomes, cancer progression, surgical risk, and healthy longevity into a single computable map. The result is not a replacement for RCTs, but a hypothesis-generating operating system for powering the next-generation of innovation.
"These studies from nference signal that longitudinal, AI-curated evidence can inform prospective studies in profound ways," said Dr. C. Michael Gibson, MD, Professor of Medicine at Harvard and CEO of the Baim Institute for Clinical Research. "The broad spectrum of weight-loss-independent benefits revealed by the first series of MAIA publications challenges long-held cardiometabolism assumptions that place glucose lowering and total body weight reduction at the center of therapeutic value. These findings reinforce that incretin biology is organ-specific, durable, and multidimensional.
The emerging blueprint for cardiometabolism trial design
The first wave of peer-reviewed MAIA publications reframes obesity pharmacotherapy from an average-weight-loss field into a response-architecture field. Across these studies, nference advances a new organizing principle for cardiometabolism: weight-change velocity is not merely an endpoint, but a longitudinal signal of metabolic memory, incretin responsiveness, behavioral context, and therapeutic durability. The findings illuminate the questions that conventional obesity trials have not been designed to answer at scale: which patients enter distinct response trajectories, how quickly those trajectories separate, what biological and behavioral signals precede them, and which patients continue to benefit after GLP-1/GIP receptor agonist treatment stops.
In the peer-reviewed study "Decoding the hallmarks of GLP-1RA weight-loss super-responders," published in Biology Methods and Protocols by Oxford University Press1, the MAIA team analyzed 135,349 individuals treated with semaglutide and tirzepatide formulations and classified patients into weight-loss super-responders, moderate responders, minimal-weight-loss patients, and weight regainers. Most provocatively, among super-responders, one year of treatment returned average body weight to levels approximating 20 years before therapy initiation; moderate responders approximated a 10-year reversal. This finding suggests that GLP-1RA response is not merely a treatment effect measured at one year, but a window into decades of accumulated metabolic history.
In the peer-reviewed study "Weight Loss Dynamics and Health Burden Changes with Tirzepatide versus Semaglutide," accepted by the journal PNAS Nexus (published by U.S. National Academy of Sciences)2, the MAIA team compared 10,339 tirzepatide-treated and 10,339 semaglutide-treated propensity-matched patients. An important trial-design signal unearthed was heterogeneity in weight-loss response: female and white patients on either tirzepatide or semaglutide more frequently achieved significant weight loss than male, black, and Hispanic patients, who were more likely to lose less than 5% of their body weight in a year. These differences in weight-loss response outcomes for both the currently marketed GLP1RA therapies highlights the need for the next-generation of clinical studies to advance obesity solutions that cater to the broader-spectrum of patients with continued clinical unmet needs.
In the peer-reviewed study "Weight trajectories after last tirzepatide or semaglutide prescription across a federated health network" published by the journal Biology Methods and Protocols (Oxford University Press)3, the MAIA team assessed 4,182 patients in the six months after their last documented semaglutide or tirzepatide prescription. Contrary to the assumption that GLP-1RA cessation inevitably leads to immediate rebound, approximately two-thirds of patients showed stable weight or continued weight loss. In an LLM-curated subset with clinician-documented discontinuation, 72% did not show the expected weight regain. Exercise counseling appeared significantly more frequently among patients with durable weight loss than among those with regain, suggesting that post-GLP-1RA durability may be shaped by a combination of pharmacology, patient behavior, clinical follow-up, and care-delivery context.
Together, these studies define a new blueprint for cardiometabolism development: enrich trials by response phenotype, measure weight velocity rather than only endpoint weight change, treat durability after discontinuation as a first-class endpoint, and design maintenance and tapering strategies prospectively.
"This may be the first time that decades of body-weight history have been reconstructed from routine patient records at this scale," said Dr. Marcelo Montorzi, MD, an Internal Medicine physician scientist at Southcoast Health. "The finding that GLP-1 super-responders are often the same patients who experienced the most recent acceleration in weight gain is a fundamental observation. It suggests that future cardiometabolism trials should not only ask how much weight a drug reduces, but whether the patient's prior weight-gain velocity predicts the magnitude, durability, and clinical meaning of response."
From Weight Loss to Longevity: The Whole-Body Biology of GLP-1 Therapies
Across cardiovascular, hepatic, neuropsychiatric, neurodegenerative, renal, and infection-resilience studies, nference identified associations suggesting that dose exposure, treatment trajectory, and tissue-specific outcomes may provide additional insight into patient benefit. Collectively, these findings support a next-generation trial-design framework for the incretin era — one that evaluates GLP-1 biology across organs, exposure patterns, and longitudinal disease trajectories, not solely through the magnitude of weight loss.
In the study "Semaglutide cardiovascular outcomes align more closely with attained dose than achieved weight loss," published as a preprint with peer-review ongoing at npj Cardiovascular Health (Springer Nature)4, the MAIA team analyzed 47,199 patients with baseline cardiovascular disease. Higher attained semaglutide dose was associated with lower post-landmark risk of all-cause mortality, composite cardiovascular events, cerebrovascular disease, heart failure, and valvular or rheumatic heart disease. In contrast, achieved weight loss did not show the same monotonic association with later cardiovascular risk. Notably, many patients had their peak semaglutide exposure during the two years before the subsequent cardiovascular outcome window, raising a provocative hypothesis: the heart may retain a therapeutic benefit legacy of prior incretin exposure even after active treatment stops.
"The weighing scale may be the loudest biomarker of GLP-1 therapy effectiveness, but the heart may be reading a distinct signal of therapeutic benefit," said Dr. Kalyanam Shivkumar, MD PhD, Distinguished Professor of Cardiology, Radiology, and Bioengineering at the University of California, Los Angeles (UCLA). "If confirmed prospectively, this finding argues that cardiovascular trials should treat dose exposure, persistence, and organ endpoints as first-class variables rather than assuming weight loss is the major mediator."
In the study "Semaglutide is associated with stiffness improvement and broad liver benefits with distinct dose- and weight-linked patterns", the MAIA team analyzed 6,734 patients with baseline liver disease burden and a complementary elastography cohort of 326 adults with paired noninvasive liver stiffness measurements from radiology imaging before and after semaglutide initiation5. Median liver stiffness decreased after semaglutide initiation; 59.5% of patients had lower follow-up stiffness, 40.8% achieved at least 20% stiffness reduction, and 21.2% shifted to a lower fibrosis stage by prespecified elastography thresholds. These liver imaging-based findings point to potential tissue-level remodeling benefits of semaglutide. While prospective validation is needed, the results support inclusion of organ-specific endpoints in future incretin trials and position liver health as an important dimension of whole-body cardiometabolic benefit.
In the study "Legacy neuropsychiatric benefit after semaglutide is linked to maximum achieved dose and independent of the maximum weight lost", the MAIA team evaluated 63,215 semaglutide-treated patients with baseline neuropsychiatric conditions across 24 incident outcomes6. Higher attained dose during the first two years of semaglutide therapy, but not body weight lost during that period, was associated with significantly lower subsequent 2–4-year incidence of several neuropsychiatric outcome groups. The semaglutide-associated incident risk reduction includes substance disorders, mood-, anxiety- and stress-related conditions, central nervous system (CNS) atrophy, neuromuscular diseases, sleep disturbances, behavioral disorders, and impulse control diseases. By extending beyond one or two neuropsychiatric endpoints, this MAIA study suggests that semaglutide's long-term neuropsychiatric associations may reflect more than weight reduction alone. In patients with prior neuropsychiatric burden, attained dose appeared more informative than maximum weight lost for subsequent incident risk across a broad spectrum of brain and behavior outcomes.
In the study "Functional-vs-Cognitive Decline Heterogeneity in Alzheimer's Disease Revealed by LLM-curation of Decades of Clinical Notes has Implications for Semaglutide Trial Design", the MAIA team recovered clinical outcome scores from decades of routine clinical notes7. Among 131,824 patients with structured Alzheimer's disease diagnoses, 42,242 had at least one LLM-derived outcome assessment, and semaglutide-treated AD patients showed more favorable paired MMSE and MoCA changes than matched controls. This study advances a trajectory-aware framework for Alzheimer's disease trial design by showing that decades of routine clinical notes can be transformed into standardized cognitive, functional, global severity, and neuropsychiatric trajectories at scale. By revealing distinct patterns of functional-first decline versus cognitive-first decline, and by linking semaglutide exposure to more favorable observed MMSE and MoCA trajectories versus matched controls, the study identifies disease phase, functional reserve, multimorbidity, APOE context, and the temporal ordering of decline as variables that may be critical for designing future incretin trials in Alzheimer's disease.
In the study "Semaglutide use linked to lower COVID-19 and influenza severity with better cardiovascular, pulmonary, renal outcomes after pandemic or seasonal infection", the MAIA team compared 11,238 COVID-19 patients and 3,409 influenza infection (flu) patients with their respective matched standard-of-care cohorts8. Prior semaglutide exposure was associated with lower 30-day COVID-19 mortality, hospitalization, ICU admission, mechanical ventilation, and composite severity, and with lower influenza hospitalization, ICU admission, mechanical ventilation, and composite severity. The study also reported lower acute coronary syndrome, pulmonary complication, and renal complication signals during sub-acute post-infection windows.
"These findings raise a host-resilience hypothesis," said Dr. AJ Venkatakrishnan, PhD, Senior Vice President for Strategic AI Partnerships at nference. "The next generation of incretin studies may focus not only on chronic disease, but also evaluate how incretin interventions could reshape how vulnerable hosts respond to acute infection, organ stress, and inflammatory injury."
Together, these MAIA studies move GLP-1 medicine from a weight-loss paradigm toward a Whole-Body Spatial Intelligence framework. The emerging question is whether the heart, liver, brain, lung, kidney, immune system, and functional reserve each carry distinct, measurable signatures of incretin exposure — and whether those signatures can be prospectively tested to advance healthy longevity. The MAIA framework now extends naturally to complex, chronic conditions with immunologic dysregulation mechanisms and substantial unmet need, including cancer and type 1 diabetes, where metabolic state intersects in intricate ways with the human immune system.
Cancer Metabotypes: Extending MAIA from Breast Cancer Survival Signals to Solid Cancer Immunotherapy and Hematologic Malignancies
In the study "Semaglutide is associated with improved breast cancer survival, lower metastatic burden, and a dose–survival relationship uncoupled from weight-loss magnitude", the MAIA team analyzed real-world outcomes from large cohorts with multimodal clinical data9. Semaglutide exposure was associated with more favorable overall survival, lower metastatic burden, and a dose-associated survival gradient in breast cancer that was not explained by weight-loss magnitude alone. Across multiple cancer types, GLP-1 receptor agonist exposure was associated with lower rates of metastatic progression and more favorable cancer trajectories. These findings raise the possibility that incretin therapies may influence cancer biology beyond their metabolic effects and support prospective studies.
At the American Society of Clinical Oncology (ASCO) 2026 Meeting10,11, the nference MAIA team presented "Metabotypes of Pembrolizumab and Lung Cancer" and "Associations between early on-treatment weight change and overall survival across targeted therapies in chronic lymphocytic leukemia (CLL)". In the lung cancer study, early on-treatment weight gain was a favorable survival association, with each 1% increase in body weight after pembrolizumab initiation consistently associated with lower subsequent mortality risk. In the CLL study of 3,542 patients initiating targeted therapies, on-treatment weight gain was associated with improved overall survival for venetoclax and ibrutinib. These studies, published in the Journal of Clinical Oncology10,11, suggest that early body-weight trajectory may serve as a pragmatic metabotype signal across both solid-tumor immunotherapy and hematologic malignancy treatment settings.
A broader implication of these oncology studies is that body-weight trajectory may be an underused clinical signal in cancer care. Many oncology trials do not systematically decode the relationship between on-treatment weight change and long-term outcomes such as survival, metastatic progression, treatment tolerance, or supportive-care need. The MAIA findings suggest that routine weight measurements, when linked to longitudinal EHR context and treatment exposure, could become a scalable, low-burden metabotype biomarker to complement immunotherapy and targeted therapy studies, while supporting earlier risk stratification, supportive-care intervention, and value-based oncology management.
"For decades, oncology has studied tumor metabolism, but we have not fully leveraged the patient's own metabolic trajectory as a clinical signal," said Dr. Manish Kohli, MD, Presidential Endowed Chair in Cancer Research at the University of Utah-Huntsman Cancer Institute. "These cancer metabotyping studies point to a common hypothesis: host metabolism may influence cancer progression, treatment tolerance, immune response, drug efficacy, and survival in ways that are measurable in routine care. If validated prospectively, weight trajectory and metabolic context could become practical tools for stratifying patients across solid tumors and hematologic malignancies."
Type-1 Diabetes with GLP-1: From Insulin Dependence to Insulin Deintensification
Type 1 diabetes (T1D) remains an insulin-dependent autoimmune disease with substantial unmet need, particularly for adults facing high insulin requirements, weight gain, glycemic variability, and long-term cardiometabolic risk. In the study, "Semaglutide and Tirzepatide in Type 1 Diabetes: Real-World Insulin Deintensification, Cardiovascular Outcomes and Safety Assessment," the MAIA team analyzed 2,002 adults with T1D exposed to semaglutide or tirzepatide versus propensity-score matched T1D control patients without GLP-1RA exposure12. The central signal was sustained insulin deintensification: at 12 months, insulin total daily dose (TDD) declined by 22.0% with semaglutide and 19.7% with tirzepatide, compared with a 10.3% increase in matched controls. By 24 months, insulin TDD reductions widened to 27.4% and 25.9% with semaglutide and tirzepatide respectively, compared with 10.4% increase in controls.
At 12 months, HbA1c declined by 4.2% with semaglutide and 3.0% with tirzepatide versus 0.3% in controls, while body weight declined by 4.0% and 6.8%, respectively, versus 0.5% in controls. Among semaglutide-treated patients, mean 12-month insulin TDD reduction was 34.5% in those losing at least 10% of body weight versus 19.6% in those losing less than 5%, and higher semaglutide dose tiers were associated with greater TDD reduction. By contrast, tirzepatide dose escalation was not significantly associated with TDD reduction in this analysis. These findings suggest that future T1D trials should jointly model insulin dose reduction, achieved incretin dose, weight change, HbA1c, and safety signals to distinguish patients who can safely de-intensify insulin from those who require closer monitoring.
The observed safety profile was gastrointestinal-predominant, with increases in nausea or vomiting, constipation, diarrhea, decreased appetite, hypoglycemia, and acute kidney injury, while diabetic ketoacidosis, severe hypoglycemia, pancreatitis, gallbladder disease, gastroparesis, and retinopathy progression did not change significantly in the paired pre/post analysis. However, patients with greater than 30% insulin TDD reduction at 6 months had higher diabetic ketoacidosis incidence, suggesting that insulin deintensification should be supported by structured safety monitoring rather than simple dose reduction.
The MAIA study also showed that across 24 prespecified cardiovascular and renal time-to-event endpoints, semaglutide or tirzepatide exposure was associated with lower 2-year all-cause mortality and major adverse cardiovascular events (MACE), while no renal endpoint remained significant after multiple-testing correction.
"The recent publications from nference provide a new roadmap of AI-curated evidence that will almost certainly be used to inform gold standard randomized clinical trials for disorders beyond obesity and type 2 diabetes mellitus," said Dr. Clifford Rosen, MD, Director of the Center for Clinical & Translational Research at Maine Health and Professor at Tufts Medicine.
Physician-adjudicated LLM Curation of Longitudinal Clinical Notes Reveals the Hidden Landscape of Treatment adherence and Compounded GLP-1 Use
In the study "Auditable LLM curation of clinical notes refines GLP-1 initiation, persistence ascertainment and compounding use beyond prescription records," the MAIA team exposed a central vulnerability in GLP-1 real-world evidence: a prescription is not the same as a patient taking the drug13. Across 553,073 adults with semaglutide or tirzepatide prescriptions, structured prescription records were noted to miss salient information that an LLM pipeline on clinical notes captured, such as delayed starts, access barriers, early interruptions, true persistence, and compounded GLP-1 exposure. Prescription-based persistence was found to substantially overestimate note-confirmed active use at 18 months, while insurance and cost barriers, perioperative holds, and gastrointestinal adverse events emerged as major drivers of discontinuation or treatment friction. This study demonstrates that clinical notes are not ancillary evidence, but essential infrastructure for GLP-1 trial design, payer strategy, safety surveillance, compounding-risk monitoring, and interventions that improve sustained access to peptide therapies whose benefits depend on starting and staying on treatment.
"Structured prescriptions record intent, whereas longitudinal clinical notes reveal reality," said Karthik Murugadoss, Vice President of AI Research at nference. "For GLP-1 therapies, critical signals such as exposure, adherence, compounding, weight dynamics, and clinician judgment often live beyond structured EHR fields. This study shows how LLMs with physician validation turn fragmented health system encounters into auditable AI infrastructure for the incretin era."
Semaglutide Microdosing Linked to Cardiometabolic Benefit with Minimal Weight Loss; Head-to-Head Comparison Shows Favorable Tolerability Over Low-Intensity Tirzepatide
In the study "Sustained low-dose semaglutide is linked to broad-spectrum cardiometabolic benefits, and better tolerability profile over low-dose tirzepatide, motivating semaglutide microdosing studies," the MAIA team analyzed real-world outcomes among patients who remained on initiation-dose semaglutide (0.25 mg) or initiation-dose tirzepatide (2.5 mg) for prolonged periods14. Despite only modest weight change, sustained initiation-dose semaglutide was associated with lower 24-month event probabilities for mortality, composite cardiovascular events, heart failure, arrhythmias, ischemic heart disease, and venous thromboembolism versus matched patients receiving metformin, DPP-4 inhibitors, or SGLT-2 inhibitors, with the strongest signals observed among patients with baseline cardiovascular burden. In head-to-head comparisons, sustained initiation-dose semaglutide also demonstrated a more favorable tolerability profile than sustained initiation-dose tirzepatide, with lower rates of diverse gastrointestinal, renal, musculoskeletal, and respiratory adverse events. These findings suggest that clinically meaningful cardiometabolic benefits may be achievable at substantially lower doses than currently studied in RCTs and support prospective evaluation of semaglutide microdosing strategies. Particularly, low-intensity semaglutide exposure may represent a distinct opportunity for prevention, maintenance, and longevity-oriented cardiometabolism trials.
"The very promising link of microdose levels of these agents to lower 24-month event-rates for a broad array of cardiovascular endpoints despite very modest weight loss, when taken together with evidence of disproportionate lean muscle loss at higher doses of some agents, only emphasizes the reality that the magnitude and scope of the many and varied impacts of these agents are well beyond their effectiveness at overall weight loss," said Dr. Joseph Smith, MD PhD, Cardiologist/Electrophysiologist, AI interventions expert and Adjunct Professor of Johns Hopkins.
Body-composition digital phenotyping reveals that the future of GLP-1 longevity medicine must optimize fat loss without sacrificing lean-tissue reserve
In the study "Greater lean-body-mass decline with tirzepatide than semaglutide in routine care, revealed by body-composition digital phenotyping," the MAIA team showed that the next frontier in GLP-1 medicine is not simply how much weight is lost, but what kind of weight is lost15. Across 670,422 first-episode GLP-1RA users, including 456,742 semaglutide-treated and 213,680 tirzepatide-treated patients, 7,965 individuals had paired pre- and post-initiation body-composition measurements. Tirzepatide was associated with greater relative lean-body-mass (LBM) loss than semaglutide at every measured timepoint, with excess LBM losses of 1.1%, 1.5%, 1.3%, and 2.0% at 3, 6, 9, and 12 months, respectively. Tirzepatide also produced greater total body weight (TBW) loss, suggesting that its superior weight-loss efficacy may come with a greater LBM tradeoff. A "Depletive GLP-1 Metabotype," defined as >20% TBW loss with >5% LBM loss, was significantly more frequent with tirzepatide than semaglutide (10.3% versus 6.7%), while the "Prime GLP-1 Metabotype," defined as >10% TBW loss with <5% LBM loss was 11.8% versus 12.3%, respectively.
The implications are clinically important for U.S. adults with Medicare coverage (65+ years), patients with frailty risk, adults seeking longevity-oriented long-term metabolic optimization, and patients with sports injuries, chronic knee or neck pain, reduced mobility, or other conditions where preserving lean mass is central to function. Indeed, in this MAIA study15, baseline musculoskeletal pain emerged as a major correlate of greater LBM loss, with cervicalgia and knee pain showing especially large LBM loss, and higher LBM loss was linked to reduced exercise tolerance. These findings argue that future cardiometabolism trials should incorporate whole body composition-aware endpoints that maximize fat loss while preserving strength, function, and the lifelong reserve needed for the quest for longevity.
Toward Autonomous AI Agents for Healthy Longevity
The first series of MAIA publications highlights a larger quest at nference: engineering autonomous AI agents that continuously learn from de-identified routine care, convert multimodal patient trajectories into computable hypotheses, and help design the next generation of prospective cardiometabolism studies. Within MAIA, these agents operate with physician and scientist checkpoints, generating prompts, challenging and validating one another's hypotheses, and mapping incretin therapies, concomitant medicines, and real-world interventions to their molecular mechanisms of action. This enables systematic assessment of potential synergy, antagonism, and signaling-pathway interactions across routine care. MAIA represents an early step toward a future in which cardiometabolism is interpreted as a living system of biomedical knowledge, and real-world evidence becomes a translational intelligence layer for advancing AI diagnostics, precision peptide therapeutics, surgical interventions, clinical decision support, and healthy longevity.
References
AJ Venkatakrishnan, Karthik Murugadoss, Venky Soundararajan, Decoding the hallmarks of GLP-1RA weight-loss super-responders, Biology Methods and Protocols, Volume 11, Issue 1, 2026, bpag021,
AJ Venkatakrishnan, Karthik Murugadoss, Venky Soundararajan, Weight Loss Dynamics with Tirzepatide versus Semaglutide, PNAS Nexus, Accepted for publication, June 2026, medRxiv preprint:
Karthik Murugadoss, Gowtham Varma, AJ Venkatakrishnan, C Michael Gibson, Venky Soundararajan, Weight trajectories after last tirzepatide or semaglutide prescription across a federated health network, Biology Methods and Protocols, Volume 11, Issue 1, 2026, bpag020,
Karthik Murugadoss, AJ Venkatakrishnan, Christopher Gregg, Venky Soundararajan, Semaglutide cardiovascular outcomes align more closely with attained dose than achieved weight loss, npj Cardiovascular Health - under review. Preprint in Research Square (Springer Nature),
AJ Venkatakrishnan, K Purushotham, Gowtham Kumar Varma, Avinash Aman, Karthik Murugadoss, Venky Soundararajan, Semaglutide is associated with stiffness improvement and broad liver benefits with distinct dose- and weight-linked patterns. Under review. Preprint:
Karthik Murugadoss, AJ Venkatakrishnan, Venky Soundararajan, Legacy neuropsychiatric benefit after semaglutide is linked to maximum achieved dose and independent of the maximum weight lost. Under review. Preprint:
Karthik Murugadoss, AJ Venkatakrishnan, Gowtham Kumar Varma, Ravi Sunil Kumar, Gourab Saha, Venky Soundararajan, Functional-vs-Cognitive Decline Heterogeneity in Alzheimer's Disease Revealed by LLM-Curation of Decades of Clinical Notes has Implications for Semaglutide Trial Design. SSRN (Preprints with The Lancet):
Robert Matson, Karthik Murugadoss, AJ Venkatakrishnan, Venky Soundararajan. Semaglutide use linked to lower COVID-19 and influenza severity with better cardiovascular, pulmonary, renal outcomes after pandemic or seasonal infection. Submitted to Preprints.org.
Karthik Murugadoss, AJ Venkatakrishnan, Venky Soundararajan. Semaglutide is associated with improved breast cancer survival, lower metastatic burden, and a dose–survival relationship uncoupled from weight-loss magnitude. Under review. Preprint:
Venky Soundararajan, Karthik Murugadoss. Metabotyping using early body weight velocity in lung cancer patients treated with pembrolizumab. J Clin Oncol
44, e20527-e20527(2026).
Venky Soundararajan, Karthik Murugadoss. Associations between early on-treatment weight change and overall survival across targeted therapies in chronic lymphocytic leukemia. J Clin Oncol
44, e19001(2026).
Robert Matson, Karthik Murugadoss, AJ Venkatakrishnan, Venky Soundararajan. Semaglutide and Tirzepatide in Type 1 Diabetes: Real-World Insulin Deintensification, Cardiovascular Outcomes and Safety Assessment. Preprints.org.
Gowtham Varma, Karthik Murugadoss, Mathews E Kurian, Jerin Varghese, AJ Venkatakrishnan, Venky Soundararajan. Auditable LLM curation of clinical notes refines GLP-1 initiation, persistence ascertainment and compounding use beyond prescription records. Submitted to Preprints.org.
Karthik Murugadoss, AJ Venkatakrishnan, Venky Soundararajan. Sustained low-dose semaglutide is linked to broad-spectrum cardiometabolic benefits, and better tolerability profile over low-dose tirzepatide, motivating semaglutide microdosing studies. Preprints.org.
Karthik Murugadoss, AJ Venkatakrishnan, Venky Soundararajan. Greater lean-body-mass decline with tirzepatide than semaglutide in routine care, revealed by body-composition digital phenotyping. Under review. Preprint in Research Square (Springer Nature),
About MAIA
The Metabolism Agentic Intelligence Atlas (MAIA) is an nference strategic initiative focused on making cardiometabolism computable across therapies, diseases, organs, and patient trajectories. By integrating federated real-world data with AI-enabled clinical phenotyping, MAIA maps how metabolic interventions influence health outcomes across the body and throughout the lifespan. The platform combines electronic health records, clinical notes, treatment exposure, laboratory measurements, imaging-derived biomarkers, body-composition assessments, and other AI-derived phenotypes to generate biologically grounded, clinically testable insights. These insights are intended to inform clinical trial design, patient stratification, dose optimization, safety evaluation, organ-specific endpoint discovery, and AI-enabled decision support. Unless otherwise specified, MAIA studies are retrospective and observational. Associations should not be interpreted as proof of causality or used to guide clinical care without prospective validation. Preprints have not completed peer review. nference intends MAIA to serve as a hypothesis-generating translational engine that can inform rigorous prospective studies, including randomized trials, pragmatic trials, active-comparator observational analyses, mechanistic studies, diagnostic development studies, and future AI-enabled clinical decision-support frameworks.
About nference
nference is an Agentic Intelligence company making biomedical knowledge computable from multimodal healthcare data. The company builds federated AI systems that transform the language of routine care into privacy-preserving, computable knowledge for discovery, development, and care transformation. The data includes clinical notes, lab tests, medication records, radiology and pathology imaging, electrophysiology signals, body-composition data, wearables, genomes, and longitudinal patient trajectories. nference is building the infrastructure for a future in which biomedical knowledge is continuously learned from routine care, responsibly validated, and translated into better strategies for prevention, treatment, and healthy longevity.
SOURCE nference
21%
more press release views with
Request a Demo
