Translational Science Techniques

I am tremendously excited about the real-world impact of our latest publication on #AI #Biomarkers in Nature Medicine: https://lnkd.in/dv-7aS7Y Even in the US barely half of #lungcancer patients are tested for #EGFR mutations, for which targeted therapies readily exist. We have worked for many, many years now to try to overcome this gap with AI for H&E slides to offer patients a fast and cost-effective solution to get the right treatment. The point of this work is not only that we actually built it, but that Gabriele Campanella and Chad Vanderbilt organized a consortium and created the infrastructure for the first real-world, real-time deployment of a fine-tuned pathology foundation model for lung cancer biomarker detection. 𝙋𝙧𝙤𝙨𝙥𝙚𝙘𝙩𝙞𝙫𝙚𝙡𝙮!   𝐌𝐞𝐞𝐭 𝐄𝐀𝐆𝐋𝐄 (EGFR AI Genomic Lung Evaluation): ✅ 𝟎.𝟖𝟗 𝐀𝐔𝐂 in a 𝐩𝐫𝐨𝐬𝐩𝐞𝐜𝐭𝐢𝐯𝐞 silent trial with clinical-grade performance. 🌍 Generalizes 𝐚𝐜𝐫𝐨𝐬𝐬 𝐡𝐨𝐬𝐩𝐢𝐭𝐚𝐥𝐬 𝐚𝐧𝐝 𝐜𝐨𝐧𝐭𝐢𝐧𝐞𝐧𝐭𝐬 with robustness and reproducibility. 🔬 Validated on 𝐢𝐧𝐭𝐞𝐫𝐧𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐜𝐨𝐡𝐨𝐫𝐭𝐬, 𝐦𝐮𝐥𝐭𝐢𝐩𝐥𝐞 𝐢𝐧𝐬𝐭𝐢𝐭𝐮𝐭𝐢𝐨𝐧𝐬, 𝐚𝐧𝐝 𝐬𝐜𝐚𝐧𝐧𝐞𝐫𝐬. 🧪 𝟒𝟑% 𝐫𝐞𝐝𝐮𝐜𝐭𝐢𝐨𝐧 𝐢𝐧 𝐫𝐚𝐩𝐢𝐝 𝐦𝐨𝐥𝐞𝐜𝐮𝐥𝐚𝐫 𝐭𝐞𝐬𝐭𝐬, preserving biopsy tissue for full genomic profiling. ⚡ 𝐃𝐞𝐥𝐢𝐯𝐞𝐫𝐬 𝐫𝐞𝐬𝐮𝐥𝐭𝐬 𝐢𝐧 𝐮𝐧𝐝𝐞𝐫 𝟏 𝐡𝐨𝐮𝐫, compared to 2–3 weeks for NGS. 🚀 A foundational step toward regulatory approval and 𝐀𝐈-𝐢𝐧𝐭𝐞𝐠𝐫𝐚𝐭𝐞𝐝 𝐜𝐥𝐢𝐧𝐢𝐜𝐚𝐥 𝐰𝐨𝐫𝐤𝐟𝐥𝐨𝐰𝐬.   We have worked on Computational Biomarkers in Pathology continuously for over a decade starting with AI for predicting SPOP in prostate cancer from H&E in 2015, but seeing everything come to fruition at such a scale in 2025 is very humbling. AI, when done right, can give real, tangible help to cancer patients. 𝑰𝒕 𝒊𝒔 𝒐𝒖𝒓 𝒓𝒆𝒔𝒑𝒐𝒏𝒔𝒊𝒃𝒊𝒍𝒊𝒕𝒚 𝒕𝒐 𝒎𝒂𝒌𝒆 𝒊𝒕 𝒂 𝒓𝒆𝒂𝒍𝒊𝒕𝒚! I am deeply grateful to everyone on this most amazing team: Gabriele Campanella, Neeraj Kumar, Ph.D., Swaraj Nanda, Siddharth Singi, Eugene Fluder, Ricky Kwan, Silke Mühlstedt, Nicole  Pfarr, Peter Schüffler, Ida Häggström, Noora Neittaanmäki, Levent Akyürek, Alina Basnet, Tamara Jamaspishvili, Michel Nasr, Matthew Croken, Fred Hirsch, Arielle Elkrief, Helena Yu, Orly Ardon, Greg Goldgof, Meera Hameed, Jane Houldsworth, Maria E. Arcila, Chad Vanderbilt #AI #ComputationalPathology #Biomarkers #AIinHealthcare #DigitalPathology #PrecisionMedicine #LungCancer #EGFR #NatureMedicine #FoundationModels #EAGLEModel #EAGLE #Oncology

AI could make clinical trials faster, cheaper, and more inclusive, but success depends on explainability, interoperability, and trust. 1️⃣ 80% of trials face recruitment delays, and 50% of datasets contain quality issues; AI aims to fix both. 2️⃣ Machine learning improves protocol design accuracy (80% vs. 65%) and accelerates site selection and feasibility assessments. 3️⃣ AI tools boost enrollment by up to 65% and cut screening time by 78%, though real-world deployment can be costly and complex. 4️⃣ NLP and digital systems help identify underrepresented groups, supporting more diverse and inclusive recruitment. 5️⃣ AI-driven digital biomarkers enable 90% sensitivity in real-time safety monitoring, improving adverse event detection. 6️⃣ Risk-based monitoring powered by AI detects data integrity issues within 48 hours, much faster than manual reviews. 7️⃣ Predictive models achieve 85-90% accuracy in forecasting outcomes and enable adaptive, personalized trial designs. 8️⃣ High-dimensional, noisy, and heterogeneous data challenge AI systems; success requires strong data harmonization and validation. 9️⃣ Regulatory gaps, stakeholder distrust, and lack of explainability remain major barriers to clinical adoption. 🔟 Real-world trials show AI's promise, but also its high cost, customization demands, and integration hurdles. ✍🏻 David Olawade (MPH, FRSPH, FHEA), Sandra Chinaza Fidelis (RN, BNSc, MSc, MPH), Sheila Marinze, Eghosasere Egbon, Ayodele Osunmakinde, Augustus Osborne. Artificial intelligence in clinical trials: A comprehensive review of opportunities, challenges, and future directions. International Journal of Medical Informatics. 2026. DOI: 10.1016/j.ijmedinf.2025.106141

Are we realising the potential of our networks to make change happen? Most innovation emerges from collaborative projects where teams openly “borrow” & adapt each other’s (often small but powerful) ideas. Many networks & communities of practice could achieve so much more by experimenting together around collective priorities to generate & share new solutions. This is beyond spreading known “best” or “good” practices. It is about innovating to design new solutions collectively. So I appreciated this piece from Ed Morrison about three different kinds of networks: - Advocacy networks are communities that seek to mobilise people, creating pressure to shift policies, priorities or messages in a particular direction. Their aim is to connect & influence rather than to change how they themselves work. - Learning networks are communities of practice. They share knowledge, compare practice & build shared capability. Learning networks often excel at spread & improvement of existing practice, but only sometimes move into structured innovation work. - Innovating (or transforming) networks are communities that combine their assets - ideas, relationships, data, capabilities - to create new value that none could produce alone. They manage collaboration as a process of experimentation: agreeing a shared outcome, running multiple connected tests of change, learning by doing & amplifying what works across the network. https://lnkd.in/edbbexiG. Every learning network has the potential to become an innovating/transforming network. Some actions to enable this: 1. Build a foundation of strong, trusting relationships within the network, understanding each member’s starting point & motivation for change 2. Focus on helping each other to succeed; listen to each others’ stories & plans, co-coach, give advice to each other & build shared inquiry 3. Move from “sharing” or “raising awareness” to some concrete outcomes the network want to change together through collective experimentation 4. Agree some simple norms for the network so that members help each other to make progress, make it safe to try things, fail fast & share incomplete work 5. Encourage multiple, parallel tests of change around similar outcome so projects can “steal with pride” from one another & quickly refine promising ideas 6. Put simple routines in place for noticing patterns (what is shifting where & why), capturing these insights & amplifying them across the network 7. Add additional success metrics including innovations tested, adapted & adopted in multiple places Graphic by Ed Morrison. Content with added inspiration from June Holley.

Revolutionizing clinical trials is no small feat, but it's exactly what we're tackling in our new #AISTATS2025 paper! We introduce RFAN (Randomise First, Augment Next) - a novel two-stage framework that reimagines trial design, making it not only more efficient but also fairer for all patients. By integrating causal deep #Bayesian active learning, RFAN goes beyond conventional #RCTs, dynamically adapting recruitment and treatment assignment to ensure underrepresented populations are better included while maintaining regulatory rigour. Our approach optimises both Post-Trial Mean Benefit (PTMB) and Post-Trial Fairness (PTF), ensuring treatments reach the right people faster, with greater real-world impact, and without compromising trial integrity. This is the future of adaptive, equitable, and efficient clinical trials. by Omer Noy Klein, Alihan Hüyük, Ron Shamir, Uri Shalit, Mihaela van der Schaar Read more: 🔗 https://lnkd.in/gj4CkvXV For more information on our work on Clinical Trials, please find our research pages here: https://lnkd.in/eeay7Qqg https://lnkd.in/eYBVRj5D

The recent European Court of Justice ruling (C-413/23 P, 4 September 2025) may quietly become one of the most influential data-governance decisions for health research in years, and a game changer for patient evidence. The Court clarified that not all pseudonymised information is automatically “personal data.” The decisive factor is whether the recipient of the data can realistically re-identify the individual. If not, such pseudonymised data might fall outside the full scope of GDPR — as long as governance, ethics and transparency remain robust, and as long as the data controllers – who still hold the key to re-identification – implement all safeguards that data remains personal and fully subject to GDPR. This “relative identifiability” test could profoundly reshape how we handle survey data, patient experience studies, PROs and preference research, or also retrospective use of clinical trial data. For those of us who have worked in data projects like the HARMONY Alliance Foundation, this decision feels like a validation for the de-facto anonymization procedure we introduced with the involvement of patient organisations, a legal firm and an ethics committee in 2019. Our HARMONY approach - de-facto pseudonymisation, key separation, and controlled data access - anticipated precisely this logic: keeping data useful for science, yet safe and well governed, while making 100.000s of patient datasets accessible to big data analytics without the need to re-consent. The ruling also raises important questions: • What counts as “realistically possible” re-identification remains context-dependent and will require careful assessment. • Data Controllers must still comply with all GDPR obligations, including transparency about possible recipients and purposes. • The risk of re-identification can change over time as new data or technologies emerge. • Ethical considerations, public trust and the role of patient oversight remain crucial – particularly in sensitive domains such as rare diseases or genetic data. Research of Teodora Lalova-Spinks to which we contributed last year demonstrated that patient involvement in data governance increases patients’ willingness to agree to data sharing. As a bottom line, thoughtful data governance is paramount for data sharing and research use. This ruling does not open a free pass for data use – but it offers a clearer framework for sharing and analysing health data responsibly. We must design governance, transparency, and trust frameworks that ensure these new freedoms are used for patients, not at their expense. It will be important for patient organisations, researchers, ethics bodies and regulators to interpret and apply the ruling carefully in the coming months. https://lnkd.in/dSe65dDu What do you think Ernst Hafen, Teodora Lalova-Spinks?

🔬 Towards Decentralized and Privacy-Preserving Clinical Trials 🧠💡Register, learn and build Decentralization in clinical research is not just about scalability or cost-efficiency. It’s a cryptographic transformation that redefines trust and data sovereignty in medical innovation. Technologies like Zero-Knowledge Proofs (ZKPs) and Fully Homomorphic Encryption (FHE) are enabling a new paradigm in decentralized trials: ✅ Privacy without compromising verification: With ZKPs, patients can prove eligibility (inclusion/exclusion criteria) without revealing their full medical history. Compliance is validated without exposing sensitive data. ✅ Computation over encrypted data (FHE): FHE allows researchers to run statistical analyses and predictive models directly on encrypted datasets. No need to decrypt—privacy is preserved even during processing. Ideal for multicenter trials or pharmacogenomic studies. ✅ Traceability without surveillance: Combining blockchain with ZK/FHE enables immutable and auditable recording of clinical events (dosage, adverse effects, outcomes) without identifying the patient. 🌐 In this new model: Data stays where it’s generated (edge computing, patient devices) No centralized data hoarding or exposure risks GDPR and similar regulations are met by design, not workaround 📣 If you're working at the intersection of digital health, cryptography and clinical innovation, this is the future: crypto-technology powering secure, precise, and ethical research. #ZKProofs #FHE #DeSci #DecentralizedTrials #PrivacyByDesign #Web3Health #DigitalTrust #Blockchain #ClinicalResearch #HealthTech Anthony Joaquim José Daniel Dr. Hidenori Vivek Helena Lars Yousuke Carlos Iker Paris João Domingos

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

What makes a biomarker…a good biomarker? Each year I teach a course on biomarkers for clinicians. Here are my main takeaways: 🎯 A biomarker is, by FDA and EMA definition, a well-defined, measurable indicator of some biological ground truth. ➡️ the biomarker is only a proxy. This means uncertainty and variation are built into every measurement. ⚖️ Variation comes in three flavours: -      pre-analytical (sample handling) -      analytical (method accuracy) -      biological (within or between individuals). 🫸 🫷 If the normal reference range is wide (large between-subject variation), but the within variation is low, then personal reference ranges are better suited for monitoring changes than population reference ranges. 🟡 An example: A patient may easily increase significantly in their creatinine before exceeding the upper limit of normal. Not noticing an individual’s habitual average but exclusively focusing on whether results are inside or outside the reference range will lead you to miss important deviances until it may be too late. 📏 Accuracy tells us how well a test separates sick from healthy. Precision tells us how closely a biomarker’s risk estimates hits to the mark. A good biomarker needs both. ⚗️ A test that shines in a specialist clinic can flop in primary care. Always validate in the population it is meant for if you want to avoid spectrum bias. 🧮 Remember Thomas Bayes: if disease prevalence is 3%, the negative predictive value of a coin toss is 97% by default. It looks impressive, but is really not. In contrast, positive predictive values struggle in low-prevalence cohorts: Even a test with 95% sensitivity & specificity would only reach a positive predictive value of 37% when pre-test probability is 3%. 💡 Concluding remark: The value of a biomarker lies not in its p-value or AUC, but in whether it helps a clinician make a better decision for a real person. As decisions are dichotomous in nature, cut-offs remain among the top important features of any test.

💡 Why networking is important for research projects and consortia? 💊 In our 10-year research project (now in year 6) and multi-country consortia (9 partners), we have a work package - Networking, which I am co-leading. The aim is to create an environment that enables knowledge sharing and fosters long-standing research partnerships. Our Networking initiative is two-fold: - Networking within the Network: focused on enhancing inter-institutional communication, organizing joint activities, and fostering knowledge sharing and partnerships among our consortium. - Networking outside the Network: collaborating with local and global research initiatives, experts, and communities to create visibility, reputation, and connectivity. Our activities include topic-specific symposia at the forefront conferences in the field; "coffee clubs" for junior-senior scientist exchange; writing of joint (expert) opinions and statements; joint funding applications and of course social gatherings. Often overlooked, networking activities are invaluable and deserve a spotlight. Honestly, I am grateful to the funding body for recognizing its significance as a separate work package and including the milestones to report on. Here's why: 🤝 Networking enables interdisciplinary approaches, breaking silos and enriching our research endeavors. 🔍 Networking provides a platform for researchers to exchange information, share insights, support each other, and stay updated on the latest developments in their respective fields. 🗺 Engaging with researchers worldwide broadens our horizons, fosters diversity in thinking, and elevates the global impact of our research. There are countless more benefits! Share your experiences and examples. P.S. We also have a Policy work package 😉 #Research #Networking #Collaboration #Innovation