How to Use Biomarkers for Disease Prediction

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

Large-scale Plasma Proteomic Profiling Unveils Diagnostic Biomarkers and Pathways for Alzheimer's Disease A large-scale study examined nearly 7,000 plasma proteins from 1,270 people with clinical Alzheimer's disease (AD) and 2,096 cognitively normal individuals. We identified 456 significant aptamers (416 proteins) that showed consistent results in both discovery and replication stages, that were further validated using two external datasets, confirming their reliability. Among the 416 proteins identified in the study, including 168 proteins (193 aptamers) novel proteins associated with Alzheimer's disease (AD). These included proteins like SPC25, CLU, and PPBP involved in signal transduction, and SPARC, NCAM1, and VEGFA involved in endothelial pathways. Additionally, 122 proteins (123 aptamers) from this study have support from previous plasma studies. Since blood collection is minimally invasive, developing blood-based biomarkers would be ideal. Plasma ptau217 is very effective at predicting amyloid positivity but is a proxy for brain amyloidosis, not overall dementia. New anti-Aβ therapies can remove Aβ deposits as seen in amyloid PET scans, and studies show ptau217 levels decrease with amyloid removal, even if neurodegeneration continues. Therefore, additional biomarkers beyond tau and Aβ are needed to track overall disease status and learn whether they could be employed to develop innovative predictive models. In order to develop Aβ and tau-independent biomarkers, we used AI to identify seven-proteins that predicted clinical AD and biomarker status. This model was further validated in orthogonal platforms including Alamar and Olink and replicated in four independent cohorts. Further analysis of the predictive performance of this model with other non-AD dementia was examined and showed low overlap with Parkinson’s Disease (PD) and Dementia with Lewy Bodies (DLB), but not with Frontotemporal Dementia (FTD). Larger studies on these groups are needed to create disease-specific predictive models to better assess how specific the current model is for Alzheimer's disease. We performed pathway analysis of the 416 AD-associated plasma proteins provided insights into the biological mechanisms of AD. Although the enriched pathways seem to cover general processes like blood homeostasis and extracellular matrix signaling, a detailed analysis shows they also involve relevant endothelial and neuronal proteins. Key pathways included lipid metabolism, immune and hemostatic response, extracellular matrix, and neuronal signaling. These analyses show that endothelial cell dysfunction can cause blood-brain barrier issues, leading to brain proteins leaking into the blood. To read the full article, go to: https://lnkd.in/d-J-gMyn

Delighted to contribute to this new work published last week in Nature Genetics on disease prediction. If our aim is to empower prevention of disease by producing the most powerful and accurate prediction tools then it makes sense to explore any and all data that might be available cost-effectively. In this work, machine learning is used to enhance UK Biobank genomic and medical data with biomarkers, some collected more than a decade before the onset of disease. These 67 blood and urine markers were supplemented in secondary analyses with thousands of proteomic markers. Prediction was impressive! For example, AUC was > 0.9 for 121 ICD10 codes. In most cases, the tool outperformed stand alone polygenic risk scores whereas in cases like breast cancer and prostate cancer where PRS are particularly powerful, the standalone PRS did better. Of course, the PRS literature also clearly shows that any and all data should be integrated for best prediction. The question as always is: what is the incremental cost-value of acquiring the new data? Finally, another benefit of identifying a disease early is you can expand your definition of the disease to enhance discovery of mechanism via genome-wide or phenome-wide association. https://lnkd.in/gQN4j63u Thanks and congrats to the amazing team led by Slavé Petrovski & Dimitrios Vitsios.