Predictive Modeling Applications in Medicine

Harvard and Roche just developed a foundation AI model that predicts immunotherapy outcomes across cancers and treatments and explains why some patients respond while others don’t. Predicting who will benefit from immune checkpoint inhibitors (ICIs) has been notoriously difficult, as biomarkers like PD-L1 expression and tumor mutational burden often fail across cancer types. 𝗖𝗢𝗠𝗣𝗔𝗦𝗦 𝗶𝘀 𝘁𝗵𝗲 𝗳𝗶𝗿𝘀𝘁 𝗰𝗹𝗶𝗻𝗶𝗰𝗮𝗹𝗹𝘆 𝗴𝗲𝗻𝗲𝗿𝗮𝗹𝗶𝘇𝗮𝗯𝗹𝗲, 𝗶𝗻𝘁𝗲𝗿𝗽𝗿𝗲𝘁𝗮𝗯𝗹𝗲 𝗳𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗔𝗜 𝗺𝗼𝗱𝗲𝗹 𝗳𝗼𝗿 𝗽𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝗻𝗴 𝗶𝗺𝗺𝘂𝗻𝗼𝘁𝗵𝗲𝗿𝗮𝗽𝘆 𝗿𝗲𝘀𝗽𝗼𝗻𝘀𝗲 𝗮𝗰𝗿𝗼𝘀𝘀 𝟯𝟯 𝗰𝗮𝗻𝗰𝗲𝗿 𝘁𝘆𝗽𝗲𝘀.  1. Trained on 10,184 tumors and fine-tuned on 16 clinical cohorts spanning seven cancers and six ICI therapies, outperforming 22 baseline methods.  2. Increased precision by 8.5%, MCC by 12.3%, and AUPRC by 15.7% over the best competing models, even in new, unseen cancer types.  3. Predicted survival outcomes more accurately than PD-L1 expression and TMB, achieving a hazard ratio of 4.7 (p < 0.0001) in a phase II urothelial cancer trial.  4. Identified distinct resistance mechanisms in immune-inflamed non-responders, including TGF-β signaling, vascular exclusion, CD4+ T cell dysfunction, and B cell deficiency. A main focus of this paper is biological interpretability, something I am a huge advocate of in large models. It integrates mechanistic interpretability (concept bottleneck) with transfer learning to do so! Also to deal with uncertainty quantification beyond the learned temperature parameter, I think incorporating conformal prediction or Bayesian calibration could strengthen clinical alignment by flagging low-confidence predictions. Here's the awesome work: https://lnkd.in/gzXSnBd8 Congrats to Wanxiang Shen, Thinh Nguyen, Michelle L., Yepeng Huang, Intae Moon, Nitya Nair, Daniel Marbach, and Marinka Zitnik! I post my takes on the latest developments in health AI – 𝗰𝗼𝗻𝗻𝗲𝗰𝘁 𝘄𝗶𝘁𝗵 𝗺𝗲 𝘁𝗼 𝘀𝘁𝗮𝘆 𝘂𝗽𝗱𝗮𝘁𝗲𝗱! Also, check out my health AI blog here: https://lnkd.in/g3nrQFxW

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

Exciting news from Cambridge! Researchers have developed an AI tool that predicts if early dementia symptoms will progress to Alzheimer's with 82% accuracy. This tool uses routine cognitive tests and MRI scans, making expensive and invasive tests like PET scans unnecessary. Dementia affects over 55 million people worldwide, with Alzheimer's causing 60-80% of cases. Early detection is key, but often inaccurate without costly tests. This new AI model, developed by the University of Cambridge, changes that by using routine data to predict Alzheimer's progression more accurately than current methods. The AI categorizes patients into three groups: stable symptoms, slow progression, and rapid progression. This helps doctors tailor treatments and monitor patients effectively, enabling early interventions like lifestyle changes or new medicines. By analyzing data from over 1,900 individuals across the US, UK, and Singapore, the model predicts not only whether symptoms will progress but also the speed of this progression. It not only improves Alzheimer's care but also aims to tackle other dementias using varied data. This model's real-world applicability has been validated through independent data, showing its potential for widespread clinical use. Researchers aim to expand this tool to cover other forms of dementia and incorporate additional data types like blood markers. As we face the growing challenge of dementia, such innovations in AI offers a more accurate, non-invasive, and cost-effective diagnostic tool, vastly improving patient outcomes and healthcare resource allocation. #AI #HealthcareInnovation #Alzheimers #DementiaCare #CambridgeResearch #MedicalAI

Explainable AI is essential for precision medicine—but here's what many are missing My latest blog post unpacks a fascinating Nature Cancer paper from showing multimodal AI outperforming traditional clinical tools by up to 34% in predicting outcomes. What surprised me most? Elevated C-reactive protein—typically a concerning marker—actually indicates LOWER risk when combined with high platelet counts. Some physicians may do this in their heads but they simply cannot do this same analysis across thousands of variables systematically.  With the right multimodal data and AI systems, we can create a fundamental shift in how we develop therapies and treat patients. Here's the twist: many argue we need randomized trials before implementing these AI tools. But that’s the wrong framework entirely. Google Maps doesn't drive your car—it gives you better navigation. Similarly, clinical AI doesn't treat patients—it reveals biological patterns that already exist. The real question: Can we afford to ignore these multimodal patterns and connections in precision medicine? Or should we use AI as a tool to uncover them and help inform our decision making? Read my full analysis here: https://lnkd.in/gGA4KTip -- I'd love to hear from others working at this intersection: How is your organization approaching multimodal data integration in precision medicine? #PrecisionMedicine #HealthCareAI #CancerCare

Artificial intelligence detects breast cancer 5 years before it develops! Researchers developed an AI model called Mirai that can predict a woman's 5-year breast cancer risk from her mammogram with significantly better accuracy than current clinical methods. • Mirai outperformed both human experts and existing risk models, identifying 41.5% of patients who would develop cancer within 5 years as high-risk, compared to just 22.9% for the standard Tyrer-Cuzick model. • The AI maintained high performance across diverse patient populations in the US, Sweden, and Taiwan - addressing a key challenge for medical AI systems. • Mirai could enable more personalized screening strategies, potentially catching cancers earlier while reducing unnecessary tests for low-risk women. • The model considers both imaging features and clinical risk factors, but interestingly, the mammogram itself was far more predictive than factors like age or family history. This research highlights AI's potential to transform cancer screening and prevention. While further validation is needed, it's an exciting glimpse at how machine learning could enhance medical decision-making and improve patient outcomes. Read the full study in the comments. #AI #Healthcare #BreastCancer #MachineLearning"

New AI tool can help select the most suitable treatment for cancer patients: 🧬The tool DeepPT developed by the National Cancer Institute (NCI) in America and Pangea Biomed works by predicting a patient's messenger RNA (mRNA) profile 🧬This mRNA - essential for protein production - is also the key molecular information for personalised cancer medicine 🧬Impressively, if the tool predicted that a patient would respond to a given therapy, they would be Two to Five times more likely to respond than a patient who was predicted not to respond to it 🧬The inputs are histopathology images, essentially stained slides of patient tumour tissue, which are routinely available, cheap and fast to process - reducing delays associated with traditional molecular data processing. 🧬DeepPT was trained on over 5,500 patients across 16 prevalent cancer types, including breast, lung, head and neck, cervical and pancreatic cancers 👇Link to articles and study in comments #digitalhealth #AI