Predictive Modeling Applications in Medicine

Deep learning applied to liver MRI scan data can predict the development of cardiovascular disease. That sounds unusual, if not unbelievable, at first glance, doesn't it? In a study published online by JHEP Reports last week (—> https://lnkd.in/eP26rwDs ), Dr. Jakob Nikolas Kather from the Medical Oncology Department of the NCT Heidelberg - Nationales Centrum für Tumorerkrankungen (NCT) Heidelberg and colleagues investigated the application of transformer neural networks using liver MRI data from the U.K.Biobank’s collection to determine their efficacy in cardiovascular risk prediction. Cardiovascular disease is frequently linked to underlying metabolic conditions. Since the liver plays a central role in metabolism, it could serve as a marker for metabolic shifts that precede cardiovascular disease - particularly major adverse cardiac events (MACEs). Developing noninvasive, imaging-based biomarkers to assess cardiovascular risk, especially in individuals who have not yet shown symptoms, could support earlier detection; however, this approach remains difficult to implement. The team used transformer neural networks, a newer, more flexible type of neural network, to develop a liver-MRI foundation model trained through self-supervised learning on 44,672 U.K. Biobank single-slice liver MRIs. Of these scans, those used for the training included a combination of all of those with a recorded occurrence of MACE before the liver MRI exam (974), along with the majority of participants with no history of MACE before the MRI (43,698). An additional 750 (all 214 participants with first-time MACE after the MRI, and 536 randomly selected participants with no history of MACE both before and after the MRI) were used for external validation. In all, there were 45,422 participants. The researchers assessed the predictive ability of the model by comparing predicted risk scores with the actual cardiovascular outcomes. The team evaluated the results of subgroups based on identified risk factors from SCORE2 (e.g., diabetes, cholesterol, systolic blood pressure, sex, and smoking status) within our model’s prediction scores to “provide insight into which cardiovascular risk factors are being captured in a more pronounced way by our model.” The results showed that the model has “significant discriminatory capacity” for predicting MACE and cardiovascular-related mortality, even outperforming methods such as SCORE2. Nevertheless, the authors acknowledged that despite the model’s potential as an imaging-based biomarker for cardiovascular risk, using MRI broadly for screening is unrealistic due to its high cost and limited accessibility. Instead, they recommended its use be focused on high-risk populations or in cases where relevant imaging data already exists such as in patients with known metabolic disorders or those who have undergone liver imaging for other clinical reasons.

Hot off our recent transformer paper, we're excited to share another AI model for precision medicine! Biological data collected from patients has exploded in recent years, presenting a challenge: how do we decipher that data to understand which patients will benefit most from specific therapies?  We in the Applied Data Science team at AstraZeneca are thrilled to share our paper in Cancer Cell called "AI-Driven Predictive Biomarker Discovery with Contrastive Learning to Improve Clinical Trial Outcomes." Here, we introduce the *Predictive Biomarker Modeling Framework (PBMF)*, a neural network-powered contrastive learning process that: 🔍 Explores vast multimodal datasets to uncover predictive biomarkers in an automated, systematic, and unbiased manner  🧠 Distinguishes predictive biomarkers (which indicate a likely benefit from a specific therapy) from prognostic biomarkers (which indicate general disease outlook)  💡 Distills its outputs into an interpretable decision tree, showing what drives treatment response In our studies, the PBMF:  📊 Surpassed existing methods in finding predictive biomarkers for immunotherapy success across various cancers in clinical trial and real-world data  📈 Discovered a predictive biomarker in an early-stage trial that boosted efficacy by 15% when retrospectively applied to the corresponding phase 3 clinical trial  📈 Discovered predictive biomarkers in single-arm early phase trial data with synthetic control arms, retrospectively improving the efficacy of the corresponding phase 3 trials by at least 10% We believe the PBMF has the potential to improve the way we design clinical trials and match patients to the right therapies. It can integrate with other models like our Clinical Transformer, creating exciting possibilities to someday discover biomarkers of adverse events, dosing strategies, and even to back-translate new drug targets. Read the full paper here: https://lnkd.in/eveAnVRY   Thanks to all the co-authors: Gustavo Arango, Damian Bikiel, Gerald Sun, Elly Kipkogei, Kaitlin Smith, Sebastian Carrasco Pro, Elizabeth Choe #PrecisionMedicine #ClinicalTrials #AIinHealthcare #Biomarkers #Immunotherapy

AI trained on 400,000 health records can model how multiple diseases progress across a lifetime, sometimes matching or beating standard risk tools. 1️⃣ Delphi-2M, a GPT-style model, predicted over 1,000 diseases with accuracy similar to CVD and dementia scores, and better for mortality. 2️⃣ It generated realistic health trajectories up to 20 years ahead and created synthetic patient data without exposing real identities. 3️⃣ Cancers drove long-term mortality risk, while heart attacks and sepsis risks faded within 5 years. 4️⃣ Biases in the UK Biobank (younger, healthier, less diverse) carried into the model, showing limits of training data. ✍🏻 Artem Shmatko, Alexander Wolfgang Jung, Kumar Gaurav, Søren Brunak, Laust Hvas Mortensen, Ewan Birney, Tom Fitzgerald, Moritz Gerstung. Learning the natural history of human disease with generative transformers. Nature. 2025. DOI: 10.1038/s41586-025-09529-3

Recent advances in machine learning inform precision medicine and translational research. We introduce a pioneering approach that integrates pathology imaging with transcriptomics and proteomics to identify predictive histology features associated with critical clinical outcomes in cancer. We utilize 2,755 H&E-stained histopathological slides from 657 patients across 6 cancer types from CPTAC. Our models effectively recapitulate distinctions readily made by human pathologists: tumor vs. normal (AUROC = 0.995) and tissue-of-origin (AUROC = 0.979). We further investigate predictive power on tasks not normally performed from H&E alone, including TP53 prediction and pathologic stage. Importantly, we describe predictive morphologies not previously utilized in a clinical setting. The incorporation of transcriptomics and proteomics identifies pathway-level signatures and cellular processes driving predictive histology features. Model generalizability and interpretability is confirmed using TCGA. We propose a classification system for these tasks, and suggest potential clinical applications for this integrated human and machine learning approach. A publicly available web-based platform implements these models. Interesting paper from the Clinical Proteomic Tumor Analysis Consortium: https://lnkd.in/eAArJwDv

Hot off the press: Imagine being able to use the same kind of model as ChatGPT or Gemini to predict not the next word, but the next clinical event in a person's life. A groundbreaking new paper in Nature, "Learning the natural history of human disease with generative transformers," introduces Delphi-2M, a powerful new application of the GPT transformer architecture applied to the UK Biobank data to model the health course of volunteers, like narrative or story, and trains models the figure out how to predict the next part of the narrative of health for a different person. The model was trained on an incredible scale, using data from over 400,000 participants in the UK Biobank, a world-class population-scale research cohort, to train and then from another 100k of those participants to test and optimize hyper parameters. What’s truly exciting is how the model was then validated on a completely different population: 1.9 million Danish individuals. The implications of this work are immense. While the accuracy at the individual level may not be accurate enough for routine clinical decision making, aggregated at a population level, Delphi-2M could project the expected disease burden at local, regional, and national levels, helping healthcare systems prepare for future needs and guiding health prevention policies and strategies. And the opportunities to identify disease associations is huge—eg Figure 4B highlights some of the common conditions preceding a diagnosis of pancreatic cancer and preceding death for example. Some of these associations are known, and some are detectable only with AI. While the paper highlights the power of this approach, it also transparently discusses limitations and biases. The UK Biobank, like many research cohorts, has a "healthy volunteer bias," and its participants are not representative of the entire population, tending to be more affluent and educated. There's much opportunity to use these foundation models to achieve the goals of precision medicine, if we could get these data together. It’s also a great example about how important it is for us to understand the fundamentals of how these foundation models work so we can adapt them meaningfully to important use cases in clinical care. Kudos to the authors. Artem Shmatko Alexander Yung , Kumar Gaurav, Søren Brunak, Laust Hvas Mortensen, Ewan Birney, Tom Fitsgerald, Moritz Gerstung. EMBL European Molecular Biology Laboratory, University of Copenhagen (Københavns Universitet), Novo Nordisk Foundation foundatio  #AIinHealthcare #PrecisionMedicine #DigitalHealth https://lnkd.in/eGDRptp5

🔬 AI-Powered Prediction of Immunotherapy Response Using Routine Blood Tests from MSKCC Determining which cancer patients will benefit from immune checkpoint inhibitors (ICIs) remains a critical challenge. A new study in Nature Medicine explores SCORPIO, a machine learning model that predicts ICI efficacy using only routine blood tests (CBC, metabolic panels) and clinical characteristics—eliminating the need for complex genomic or immunologic assays. Trained on data from 9,745 ICI-treated patients across 21 cancer types and validated using over 10 global phase 3 trials, SCORPIO represents a step toward more accessible, data-driven decision-making in immunotherapy. 📖 Read more: Link to Paper in Comments 👇

We’re drowning in health data, but starving for wisdom. This may be the bridge between the two: Every year, the amount of biomarker data available to patients multiples: Wearables, multi-omics panels, continuous monitors, AI-driven dashboards. And yet, many clinicians and patients are left more confused than ever. At my clinics, I’ve witnessed a shift. When we wrap those layers of data inside AI-driven interpretation, we stop treating lab values as static snapshots and begin treating them as evolving narratives. Here’s what it means for longevity medicine: 1. AI is uncovering hidden biomarker patterns A recent review covering 236 AI-driven biomarker research articles shows how machine learning is pulling clear signals out of what would otherwise be noise. For example, in Alzheimer’s research, AI-based multi-omics integration has identified novel hub genes (e.g. APP, SOD1) linking metabolism and synaptic pathways years before clinical onset. These are precisely the kinds of signals we want earlier in longevity medicine. 2. Multimodal models outperform single-axis tests In oncology, AI models that merge imaging, lab values, clinical notes, and patient metadata are now more accurate than traditional biomarker panels in detecting cachexia or predicting survival. In a recent pancreatic cancer cohort, a multimodal AI-driven biomarker model achieved up to 85% accuracy in early detection compared with conventional methods. 3. Digital biomarkers are emerging from routine tracking A novel model using photoplethysmography (PPG) data from wearables estimated “vascular age” with startling predictive power. Individuals whose AI-vascular age exceeded chronological age by ≥9 years were at significantly higher risk of cardiovascular events, diabetes, and death.That means your daily pulse waveform (something many of us already measure) could offer deep insight when processed through the right algorithm. What this means in practice: ✅ Protocols become adaptive: dose, timing, combination therapies adjust in response to longitudinal data. ✅ We reduce “one-size-fits-all” risk: AI helps us stratify who truly benefits from which interventions. ✅ We can detect early inflection points before symptoms and intervene sooner. ✅ Patients feel seen: not as snapshots, but living systems moving through time. Of course, challenges still remain such as algorithm transparency, validation across diverse populations, data privacy and security, and regulatory pathways. Yet the momentum is clear. The domain of AI-interpreted biomarkers is becoming a clinical imperative. Where do you think AI will have the greatest impact on healthcare: research, or the clinic?

NotebookLM: "Researchers have developed a generative transformer model named Delphi to predict the complex progression of human diseases over an individual's lifetime. By adapting the architecture used for large language models, the system treats medical diagnoses and life events as sequential "tokens" to forecast future multimorbidity risks and mortality. The model was trained on extensive UK Biobank data and validated using Danish registries, demonstrating a high capacity to simulate long-term health trajectories based on past clinical history and lifestyle factors. Beyond individual risk assessment, the technology can generate synthetic health data to protect patient privacy while maintaining statistical accuracy for medical research. It also features explainable AI components that reveal how specific past illnesses, such as digestive disorders, quantitatively increase the likelihood of future conditions like pancreatic cancer. Ultimately, this framework aims to support clinical decision-making and help policymakers project the future disease burden of aging populations." From the cited source: "Here we present Delphi-2M—a GPT-based model of multi-disease progression. Delphi-2M extends the GPT large language model to account for the temporal nature of health trajectories. Analogous to LLMs, which learn the grammar and contextual logic of language from large bodies of text, Delphi-2M inferred the patterns of multi-disease progression when trained on data for more than 1,000 diseases and baseline health information recorded in 402,799 UK Biobank participants. A detailed assessment of Delphi-2M’s predictions showed that they consistently recapitulate the patterns of disease occurrence at the population scale as recorded in the UK Biobank. For the majority of diseases, Delphi-2M’s multi-disease, continuous-time model predicted future rates at comparable or better accuracy than established single-disease risk models, alternative machine learning frameworks and blood-biomarker-based models." https://lnkd.in/eURYKDPW https://lnkd.in/erQvS5Fq listen: https://lnkd.in/eepfHKXM

Posted with informed patient consent. This surgical content is shared solely for educational purposes.  This operation showcases the removal of an aortic aneurysm in a patient with a bicuspid valve and replacement of the ascending aorta. While this surgery is lifesaving, imagine a future where we can predict such conditions *before* they become critical. With AI-powered digital twins, we can create a virtual replica of a patient’s heart —Integrating real-time data from imaging, genetics, and lifestyle. These models can simulate disease progression, predict aneurysm formation, and even assess the timing for intervention. Could digital twins, combined with predictive analytics, revolutionize how we treat—and prevent—complex cardiovascular conditions? Let’s discuss how we move from reactive to *proactive* care. (Pictures of valve leaflets in the comments!) #AIInHealthcare #DigitalTwins #CardiacSurgery #PredictiveMedicine #BicuspidValve #AorticAneurysm #HealthcareAI #FutureOfMedicine #CardiovascularHealth #PrecisionMedicine #AIForGood #DataDrivenCare

Collaborative innovation combining AI with neuropsychology is proving to be transformative. Six research clusters show specific value and potential: 🌱 Neuroscience and Mental Health: Understanding mental health through neuroimaging and machine learning enables earlier, more precise interventions for conditions like ADHD and depression. By examining correlations in brain function, this research helps identify key markers for cognitive impairments, aiding in early diagnosis and personalized treatment plans. 🔍 Computational Modeling: Computational models simulate decision-making and cognitive markers, which are crucial for neurological conditions like epilepsy. Machine learning applied to seizure detection, for instance, offers a potential breakthrough in predicting and managing epilepsy, helping patients gain better control and care. 🧠 Cognitive Neuroscience: Studies of cognitive decline and neurodegenerative diseases, such as Alzheimer’s, benefit from reinforcement learning models that reveal patterns in brain degeneration. These insights are essential for developing strategies to slow disease progression, offering hope for more effective interventions. 💡 Cognitive Neurology and Neuropsychology: Examining cognitive functions through neuroimaging and machine learning provides deeper insights into disorders like aphasia and neurocognitive deficits. By mapping brain functions and assessing structural changes, these studies advance our understanding of how specific neurological impairments affect behavior and cognition. 💗 Neuropsychological Features: Machine learning models predict mental health outcomes and cognitive declines by analyzing attention and processing speed. This focus on prediction and prevention, especially for conditions like cardiovascular disease impacting cognition, enables proactive care and lifestyle adjustments to mitigate risks. ⚙️ Neurodegenerative Conditions: AI-based predictive models for neurodegenerative diseases like Parkinson’s allow for early, more accurate diagnoses. By analyzing markers in social cognition and emotional processing, this cluster supports personalized interventions, helping to maintain patient quality of life and reduce care burdens. This is only the beginning. This field is absolutely ripe for rapid advance and massive real-world value.