How to Use Biomarkers for Disease Prediction

A single measurement in midlife can tell us much more about future heart attack risk than many still assume. In this JACC paper, a model based on just four biomarkers: 1) CAD polygenic risk score 2) hsCRP 3) LDL-C 4) and Lp(a) strongly predicted future coronary artery disease incidence across men and women and across ages 40 to 69. Importantly, this biomarker-based approach performed better than traditional clinical risk calculators, while also identifying people who may be missed by conventional risk assessment. What makes this especially important is that two of these markers, PRS and Lp(a), are fundamentally different from the others. Both are stable across life and can provide a long-term baseline of inherited cardiovascular risk. Lp(a) is genetically determined and, unlike many conventional biomarkers, does not meaningfully fluctuate in the way LDL-C or inflammatory markers can. PRS is even more distinct: it is a lifelong measure of inherited susceptibility that can be assessed with a simple non-invasive saliva test and only needs to be measured once in a lifetime. That single result provides a baseline layer of risk information that is present long before disease becomes clinically visible. So we can act in time. This is also highly relevant clinically because both Lp(a) and PRS are now reflected in the new ACC/AHA dyslipidemia guideline framework, underscoring the shift toward earlier and personalized risk assessment. The guideline recommends at least one lifetime measurement of Lp(a), and the 2026 guideline includes PRS in cardiovascular risk assessment for the first time. This matters because LDL-C does not mean the same thing in everyone. We and others have shown that PRS modifies the impact of LDL-C on heart attack risk. In practice, this means that someone with apparently “normal” LDL-C may still be on a trajectory toward building dangerous atherosclerotic plaque if their inherited risk is high. Looking at LDL-C alone can therefore be misleading. Looking at LDL-C through the lens of PRS is much more informative. This is exactly where prevention needs to go: earlier, more personalized, and biologically grounded risk assessment. Link to the JACC study: https://lnkd.in/dZYUHKQn

Phenomenal #Proteomics #Biomarkers #TherapeuticTargets #Resource now breaking online at Cell Press | UK Biobank plasma yields comprehensive Open-Access #Proteome ↔ #Phenome #Atlas | #health | #disease | #diagnostics | Large-scale proteomics studies can refine our understanding of health and disease and enable precision medicine. Here*, the authors provide a detailed atlas of 2,920 plasma proteins linking to diseases (406 prevalent and 660 incident) and 986 health-related traits in 53,026 individuals (median follow-up: 14.8 years) from the UK Biobank, representing the most comprehensive proteome profiles to date. This atlas revealed 168,100 protein-disease associations and 554,488 protein-trait associations. Over 650 proteins were shared among at least 50 diseases, and over 1,000 showed sex and age heterogeneity. Furthermore, proteins demonstrated promising potential in disease discrimination (area under the curve [AUC] > 0.80 in 183 diseases). Finally, integrating protein quantitative trait locus data determined 474 causal proteins, providing 37 drug-repurposing opportunities and 26 promising targets with favorable safety profiles. These results provide an open-access comprehensive proteome-phenome resource (https://lnkd.in/dXGrTCbD) to help elucidate the biological mechanisms of diseases and accelerate the development of disease biomarkers, prediction models, and therapeutic targets. *https://lnkd.in/dJuTcVMg Celentyx Ltd Professor Nicholas Barnes PhD, FBPhS Omar Qureshi Catherine Brady GRAPHICAL ABSTRACT

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.

I track 12 biomarkers every 6 months. Because I've diagnosed 1,000+ dementia cases and I'm terrified of becoming one. Metabolic: A1c: 5.2% (target: <5.7%) Fasting glucose: 88 mg/dL (70-90) Fasting insulin: 4.2 μIU/mL (<6) Why: Every point A1c increase = 20-40% higher dementia risk. Insulin resistance drives neurodegeneration years before diabetes. Cardiovascular: ApoB: 60 mg/dL (<80) LDL particle number: 920 nmol/L (<1000) HDL: 61 mg/dL (>50) Triglycerides: 68 mg/dL (<100) Why: CV disease doubles dementia risk. ApoB and LDL particle number predict this better than standard cholesterol. Small dense LDL is more atherogenic. Inflammation: hs-CRP: 0.4 mg/L (<1.0) Homocysteine: 7.8 μmol/L (<10) Why: Chronic inflammation accelerates brain aging. Elevated homocysteine damages brain blood vessels. Brain-specific: Vitamin D: 59 ng/mL (40-60) Vitamin B12: 680 pg/mL (>400) TSH: 1.8 μIU/mL (1.0-2.5) Why: Low vitamin D correlates with cognitive decline. B12 deficiency mimics dementia. Thyroid dysfunction impairs cognition. Genetic: APOE: ε3/ε3 (average risk) What I do with this data: When A1c crept to 5.6%: Cut refined carbs, added 30 min daily walking. Dropped to 5.2% in 3 months. When hs-CRP was 1.2 mg/L: Adjusted or increased diet and exercise and addressed stress. Inflammation normalized. When ApoB was 110 mg/dL: Started low-dose rosuvastatin + ezetimibe. Now stable at 60 mg/dL. Pattern: Measure. Find problems early. Fix before damage. Cost: $300 every 6 months out of pocket. $600 annually. Compare to $100,000+ annual cost of dementia care. What you can do: Ask your doctor to add ApoB, hs-CRP, homocysteine to annual labs. If they won't, pay out of pocket. Worth it. Get A1c checked even if not diabetic. Check vitamin D and B12. Track trends. The mindset: I track biomarkers to catch problems 5-10 years before symptoms. That's the window where intervention works. After symptoms: managing damage. Before symptoms: preventing damage. I'd rather prevent. ⁉️ Do you track any biomarkers regularly? Which ones matter most to you? ♻️ Repost if prevention beats treatment every time 👉 Follow me (Reza Hosseini Ghomi, MD, MSE) for practical longevity strategies

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

I’ve seen 35-year-olds with the arteries of 70-year-olds, and they had no idea. That’s the scary thing about ageing. It doesn’t always show on the outside. You can look fit, go to the gym, eat “healthy”… and still be ageing far faster inside your arteries, your metabolism, and your cells. And most people don’t find out until it’s too late. That’s why prevention and longevity testing is must. Because the only way to know your real biological age is to measure it. Here are 10 markers I look at when assessing someone’s true healthspan: 1️⃣ Biological age blood panels – reveal if your body is ageing faster or slower than your years. 2️⃣ Epigenetic clocks – DNA tests that pick up cardiovascular risk years in advance. 3️⃣ Inflammation markers – going deeper than CRP, to uncover hidden vascular inflammation. 4️⃣ Lipoprotein(a) & ApoB – stronger predictors of heart disease than standard cholesterol. 5️⃣ Multi-cancer blood tests – early detection, long before symptoms. 6️⃣ Omega-3 levels – higher levels are linked with longer lifespan, yet most UK adults are deficient. 7️⃣ RDW – a simple blood count marker that predicts mortality. 8️⃣ HbA1c variability – sugar swings that silently age your arteries. 9️⃣ ST2 & Galectin-3 – markers of vascular stiffness and early heart failure risk. 🔟 AI-driven risk scores – combining genetics and labs into a personal “longevity map.” 💡 Practical tips: – Ask for ApoB and Lp(a), not just cholesterol. – Track HbA1c trends, not just one-off results. – Don’t ignore omega-3, diet or supplements. Because longevity isn’t about adding years at the end of life. It’s about protecting the years you have right now. And when I see a 35-year-old with the arteries of a 70-year-old, it’s a reminder: We can’t wait for symptoms. We need to act before disease shows up. That’s how you truly add life to years, not just years to life.

Sidra Science Highlights 2025: 🧪 Story #3 - Scaling Precision Medicine: From "N of 1" to Population Health. Our first two stories showed how we solve mysteries for individual patients (#PersonalizedMedicine). But how do we scale this approach to thousands of people (#PrecisionHealth)? To do that, we must move from studying static biomarkers (e.g. genetics) to dynamic biomarkers (multi-omics), and transition our models from #diagnostic testing to proactive risk #prediction and #screening. Today’s paper highlights a really unique application of this concept to a major global health problem: Type 1 Diabetes (#T1D). Many of us are familiar with the diagnosis of T1D and some attempts at curative treatments, but it's the rapid advances in early prediction and prevention that have gained the most momentum recently. The study, published in Nature Medicine, builds on work from Dr. Ammira Al-Shabeeb AKIL's lab, Principal Investigator and Director of the Metabolic and Mendelian Translational Research Program at Sidra Medicine. Working with a consortium of >25 institutions, the team analyzed a multi-ethnic cohort of >2,800 individuals, and: -Identified a signature of 50 circulating microRNAs (Dynamic Risk Score - #DRS) that track 'functional beta-cell loss' in real-time.  -Developed an AI-driven score which achieves an AUC of 0.84, significantly outperforming standard models predicting disease progression, and picking up beta-cell loss years before clinical onset of diabetes. -Successfully employed the model to predict not only beta-cell function #pre-diabetes, but also potential insulin needs post-transplant, validating its ability to gauge physiological stress even in transplanted patients. This study provides a biological framework for next generation screening of T1D 'in real time'!! As Qatar rolls out the #DIAMENA program, this Dynamic Risk Score allows us to move beyond simple risk prediction toward precision prevention—identifying imminent disease progression and detecting failure in transplant functionality in time to intervene. Congratulations to Ikhlak Ahmed, Dr. Ammira Al-Shabeeb AKIL, Anandwardhan Hardikar, Mugdha Joglekar and the global team! 📖 Read the paper in Nature Medicine: https://lnkd.in/gw_npU5M 🔗 Learn more about Dr. Akil’s Lab: https://lnkd.in/gb2t6BPF #SidraMedicine #PrecisionMedicine #T1D #NatureMedicine #PopulationHealth #ScienceHighlights2025 Western Sydney University Steno Diabetes Center Copenhagen University of Sydney KEM Hospital Pune Monash University Hrishikesh Hardikar, Noha Lim University of Auckland The Chinese University of Hong Kong The Westmead Institute for Medical Research Immune Tolerance Network

Clinicians have spent years guessing who will benefit from immunotherapy. But this paper introduces a machine learning-driven scoring system that dramatically improves pan-cancer prediction accuracy. The Problem: - Immunotherapy outcomes, particularly with immune checkpoint blockade (ICB), vary significantly across and within cancer types. - Existing biomarkers like PD-L1 and TMB lack universal predictive value across pan-cancer settings. - Complex tumor immune microenvironments (TIMEs) demand more robust, scalable analytic tools for accurate prediction. What the Authors Did: - Developed the iMLGAM R package, integrating machine learning, gene-pair analysis, and genetic algorithms. - Employed ensemble learning with models like Elastic Net, SVM, KNN, and Random Forest, optimized via genetic algorithms. - Validated predictive performance across multiple independent cohorts using multiomics and immune profiling techniques. Main Findings: - iMLGAM score reliably predicts ICB therapy response across pan-cancer cohorts, with lower scores linked to better outcomes. - Tumors with low iMLGAM scores show stronger immune cell infiltration, increased T cell activity, and favorable mutational signatures. - Notably, CEP55 was identified as a key gene promoting immune evasion; its knockdown reduced tumor aggressiveness and improved T cell function. - In vivo CEP55 suppression combined with anti-PD1 therapy significantly enhanced survival in mouse models. - iMLGAM outperformed 12 existing immunotherapy predictive signatures across multiple datasets. Implications for Cell & Gene Therapy - iMLGAM provides a ready-to-use, accurate scoring system for guiding personalized immunotherapy decisions. - The model's reliance on gene-pair analysis allows for platform-independent application, (enhancing clinical utility). - Integration into clinical workflows could minimize ineffective ICB treatment, reducing costs and avoiding toxicity. - CEP55 could be the next big immunotherapy target, potentially augmenting gene-targeted combination therapies. Kudos to the authors - great job! Anything else you'd add? Drop it in the comments.

🏆 GGT emerges as the holy grail of liver risk prediction? I just published a commentary in JHEP Reports on one of the most interesting population-level risk models seen in hepatology lately: the CORE score. 🩸 CORE is delightfully simple. Five inputs. All old friends: GGT, ALT, AST, age, sex. 🧮 Gives you a percentage risk of 10-year major adverse liver outcomes, and with high accuracy. As CORE is developed from the AMORIS 1985-96 cohort, I call the comment ”Old blood, new results.” https://lnkd.in/dr33RFHY What really caught my eye is how much CORE findings aligns with an emerging theme in the liver field where GGT appears again and again as a strong predictor of liver-related events: 📍 In UK Biobank where Carolin Victoria Schneider Pavel Strnad showed across in development and several external cohorts of various liver disease etiologies, that GGT (and IGF-1 interestingly) independently predicts liver-related mortality 📍 LiverPRO-F2 score from Evido Health Katrine Prier Lindvig also contains GGT, and its weight in the software is especially impactful for moderate fibrosis prediction & prognosis – indicating how inflammatory activity drives disease progression early on. 📍 Markers of portal hypertension and liver synthesis (dys)function like platelet count, sodium, albumin, INR and bilirubin holds stronger weight at the later stages, immediately before cirrhosis development or decompensation 🏋️♀️ Put together, the story is clear: a simple biochemical sign of hepatocellular injury can do heavy lifting when the outlook is at least 5-10 years. GGT summarises how liver inflammatory burden at population scale drives progression to outcomes that determine how patients feels, functions and survives. 🔗 CORE score in BMJ: https://lnkd.in/d7yaBrqY From Hannes Hagström Rickard Strandberg and colleagues at Karolinska Institutet 🔗 GGT-IGF-1 biomarker in Clin Gas Hep: https://lnkd.in/d3FbjZCh 🔗 LiverPRO paper in Lancet Gas Hep: https://lnkd.in/drRUcZ-y

1/ The biomarker test could change a patient’s life. What is it? 🧵 2/ A biomarker is a measurable signal—like a gene or protein—that helps doctors decide: Will this patient respond to treatment? 3/ In modern cancer trials, biomarkers decide more than eligibility. They guide dosing, measure benefit, and predict side effects. 4/ Biomarkers can speed up trials. If a biomarker correlates with outcome, it can serve as a surrogate—offering earlier clues before survival data rolls in. 5/ Some biomarkers have become household names in oncology. Take PD-1 and PD-L1. These immune checkpoints are critical in deciding who may benefit from anti-PD1 therapy. 6/ PD-L1, tested with IHC staining, is used as a companion diagnostic. High levels often mean better response. But here’s the catch: Some patients with low PD-L1 still respond. Some with high levels don’t. 7/ Tumor mutational burden (TMB) is another. More mutations mean more neoantigens—tiny red flags that help the immune system recognize cancer. High TMB often suggests better response to immunotherapy. 8/ Mismatch repair deficiency is also key. When tumors can’t fix their DNA, they mutate rapidly—again, flagging themselves to the immune system. MSI-high tumors are more likely to respond. 9/ Other markers—like EGFR or ALK mutations—can suggest sensitivity or resistance to targeted therapies. These are the backbone of precision oncology. 10/ Tumor-infiltrating lymphocytes (TILs) are another piece. If a tumor is already crawling with immune cells, it might just need a push to mount a full response. 11/ Sounds like we’ve got it figured out, right? Not so fast. 12/ Small studies mean small sample sizes. The more you slice the data—by gender, ethnicity, prior treatments—the less statistical power you have. False signals are everywhere. 13/ Then there’s confounding. A biomarker might look predictive—until you realize it correlates with something else, like age or comorbidities. 14/ Assay variability is another problem. PD-L1 tested in one lab might yield different results than another. The cutoff depends on the cancer type and even the testing platform. 15/ And biomarkers can change. What’s true before treatment may shift entirely after two cycles of therapy. A single snapshot might mislead. 16/ Still, biomarkers are the best compass we have in the fog of clinical trials. They make trials smarter, faster, and more personalized. 17/ But interpretation requires humility. No marker tells the full story. Context, biology, and patient history matter just as much. 18/ Key takeaways: Biomarkers guide trial design, patient selection, and treatment. PD-L1, TMB, MSI, and gene mutations lead the way in oncology. But they’re not perfect. Use with care. Validate. Re-test. Repeat. 19/ Behind every “positive” result is a human being hoping for a second chance. That’s why this matters. Get the biomarkers right—and we just might get the medicine right too.