Unlearn.AI

What can AI-generated digital twins tell us about longer-term obesity study data? That’s the question behind Unlearn’s collaboration with @StructureTx, announced today. Using anonymized data from Structure’s completed ACCESS open-label extension study of aleniglipron, we’ll estimate what patient outcomes might have been without the investigational treatment. These model-generated estimates are what we call “digital twins.” Comparing those estimates with observed outcomes may provide additional context for interpreting the study’s longer-term findings. Data from our analysis will be included in the poster “Aleniglipron in Participants Living with Overweight or Obesity: ACCESS Trial Open-Label Extension” at ObesityWeek. ➡️ Read the announcement: na2.hubs.ly/H08-25x0

🚀 Unlearn is heading to Boston for #CTAD2026, November 16–19. Alzheimer’s sponsors face big decisions at every stage of a trial: settling on eligibility and endpoints, sharpening an interim analysis, or working out what a completed study means for the next one. Digital twins help teams make each of those calls with more confidence. In a completed Phase 2 RCT in early Alzheimer’s, adding participants’ digital twins reduced the required sample size by 15%, or the control-arm size by 26%, with power maintained on CDR-SB at week 96. That analysis used PROCOVA, a method qualified by the EMA and supported by FDA guidance. Whether your study is in design, in progress, or has already read out, let’s talk at CTAD. See our Alzheimer’s work: na2.hubs.ly/H08ZdKC0 (na2.hubs.ly/H08ZdKC0) #Alzheimers #ClinicalTrials #DigitalTwins #AI #ClinicalDevelopment ——— @CTADconference

The hardest part isn’t the model. A decade of deploying #digitaltwins has taught us it’s making messy #clinicaldata usable, with a record that holds up to scrutiny. On December 7, Dr. Jon Walsh, Ian Fleming and Dr. Luis Olmos will share what we’re building for sponsors now: one foundation where historical data are unified, harmonized, and used to create and apply digital twins across every program decision. 45 minutes of discussion, plus live Q&A. 📆 Monday, December 7, 10 AM PT 🔗 Register: na2.hubs.ly/H08TbbL0

🚀 Jon Walsh, Co-founder and Chief Scientific Officer, and Luis Olmos, Senior Director of Clinical Affairs, are at ARPA-H’s SURPASS Proposers’ Day in Washington, DC today. SURPASS (Simulation-augmented, Real-time Platform Adaptive Seamless Trials) aims to replace phase-by-phase studies with continuous, adaptive trials that have simulation built into the design. We share that vision. Every clinical trial is a chain of high-stakes decisions, and digital twins, data, and AI can strengthen each one. With Unlearn, teams can simulate a trial before they commit, sharpen interim analyses with our EMA-qualified, FDA-supported PROCOVA method, and get more from completed studies so each readout informs the next design. If you’re at Proposers’ Day, come find us. We’d love to talk about where digital twins fit in the future of clinical trials. #ARPAH #SURPASS #ClinicalTrials #DigitalTwins ——— @ARPA_H

From single-study gains to portfolio decisions: a decade of learnings from deploying digital twins in clinical development. 🚀 Every clinical trial is an expensive bet and digital twins offer a better way. The question is no longer whether they belong in clinical development, but how to use them well. A decade of deploying them has taught us that the hardest part isn’t the model but making messy clinical data usable, with a record that holds up to scrutiny. On December 7, Dr. Jon Walsh (Co-founder & Chief Scientific Officer, Unlearn) sits down with Dr. Luis Olmos (Senior Director of Clinical Affairs, Unlearn) and Ian Fleming (VP of Product) to share how those lessons became what sponsors asked us to build: one foundation where historical data are unified, harmonized, and used to create and apply digital twins across every program decision. Every result traceable, every study informing the next. 45 minutes of discussion, plus live Q&A. 📆 Monday, December 7, 10 AM PT 🔗 Register: na2.hubs.ly/H08R39x0

Thank you to everyone who met with us at #ESMO2026 in Madrid last week! 🇪🇸 Many oncology indications now have several standard-of-care regimens that have never been compared head-to-head, and probably never will be. That topic came up again and again in our meetings. Trial-calibrated digital twins can make that comparison at the patient level. Interested in learning more? We’re at SITC in Phoenix this week. Let’s connect! 👋 #SITC2026 #Oncology #DigitalTwins ——— @myESMO, @sitcancer

Our Senior Director of Clinical Affairs, Luis Olmos, recently presented at the KoNECT International Conference in South Korea. 🇰🇷 His session, “𝗔𝗜 𝗮𝗻𝗱 𝗗𝗶𝗴𝗶𝘁𝗮𝗹 𝗧𝘄𝗶𝗻𝘀 𝗶𝗻 𝗖𝗹𝗶𝗻𝗶𝗰𝗮𝗹 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁: 𝗙𝗿𝗼𝗺 𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝘁𝗼 𝗥𝗲𝗴𝘂𝗹𝗮𝘁𝗼𝗿𝘆 𝗮𝗻𝗱 𝗥𝗲𝗮𝗹-𝗪𝗼𝗿𝗹𝗱 𝗜𝗺𝗽𝗹𝗲𝗺𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻,” explored how digital twins can increase statistical power in randomized trials through PROCOVA, support synthetic external controls where randomization is not feasible, and what EMA qualification of PROCOVA means for clinical development. We appreciate KoNECT for the opportunity to share our work and the attendees for a valuable discussion. #KoNECT #DigitalTwins #ClinicalTrials #ClinicalDevelopment

Back from NEALS 2026 in Clearwater, and grateful for every conversation. 🌴 We connected with ALS researchers, clinicians, and sponsors about how digital twins can help small and early-phase studies deliver clearer, more trustworthy readouts. We also presented our poster, “𝗔𝗜-𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗲𝗱 𝗗𝗶𝗴𝗶𝘁𝗮𝗹 𝗧𝘄𝗶𝗻𝘀 𝗮𝗻𝗱 𝗣𝗹𝗮𝘀𝗺𝗮 𝗡𝗳𝗟 𝗳𝗼𝗿 𝗧𝗿𝗶𝗮𝗹 𝗣𝗹𝗮𝗻𝗻𝗶𝗻𝗴 𝗮𝗻𝗱 𝗦𝘁𝗮𝘁𝗶𝘀𝘁𝗶𝗰𝗮𝗹 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆 𝗶𝗻 𝗔𝗟𝗦,” which showed how digital twin adjustment took a trial designed at 50% power up to 72% on key endpoints. Our ALS model is trained on data from more than 13,600 people with ALS and forecasts outcomes on ALSFRS-R, vital capacity, and plasma NfL. Thank you to everyone who stopped by, and to our partners across the ALS community. Missed us? Explore more of our ALS work here: na2.hubs.ly/H08JWxn0 #NEALS2026 #ALS #DigitalTwins #ClinicalTrials #AI

📣 Big shoutout to our Clinical Data Scientist, Ann-Kathrin Schalkamp, who presented “AI-Powered Digital Twins: Applications and Opportunities in Clinical Trials” at the EHDN Plenary Meeting in Kraków last week. A participant’s digital twin is a forecast of how that person would progress on placebo, built from their baseline data alone. Added to a randomized trial as a pre-specified covariate, it makes the treatment effect easier to see, using a method EMA qualified in 2022 and FDA concurred with. Ann-Kathrin showed what that means for trial planning. She took 444 real participants the model had never seen during training, selected with the entry criteria today’s disease-modifying HD trials use, and checked the model’s two-year forecasts against what actually happened to them. Sized like a current trial, a study of those 444 people could run with 64 fewer on placebo at the same power. Or keep all 444 and cut the chance of missing a drug that works from one in five to one in eight. Our Huntington’s Digital Twin Generator was trained on clinical data from 25k+ trial participants across Enroll-HD, REGISTRY, TRACK-HD, PREDICT-HD and completed trials. None of it would exist without the people and families who contributed to those studies, and the teams who built them. Thank you to EHDN and the session chairs for the invitation, and to everyone who stayed for the panel. More on digital twins in Huntington’s disease: na2.hubs.ly/H08B2YS0 #EHDN2026 #HuntingtonsDisease #DigitalTwins #ClinicalTrials

Heading to #NEALS2026 in Clearwater? Come find us at Poster #212. We’ll be presenting “AI-Generated Digital Twins and Plasma NfL for Trial Planning and Statistical Efficiency in ALS,” including how pre-specified digital twin adjustment (PROCOVA) would take a study designed at 50% power to 67% on ALSFRS-R and 72% on SVC. 📍 Poster #212 🗓️ October 29, 4:50–6:30 PM ET Stop by, or reach out to find time to connect while we’re there. #ALS #DigitalTwins #ClinicalTrials

#UnlearnerSpotlight 🚀 “I love the freedom to propose new ideas and actually see them through, and I get to work alongside incredibly talented people who I learn from every day. Most of all, I like knowing my work helps speed up clinical trials and get treatments to patients faster.”

💉 Obesity drugs now deliver more weight loss than is even desired in most cases, shifting the question from how well a drug works to how easily it can be tolerated.  In Part 2 of his obesity in clinical trials series, our co-founder and Chief Scientific Officer Jon Walsh looks at what that means for the future of obesity trials. As more treatments reach the market, dropout and drop-in (placebo patients starting marketed drugs mid-trial) will change, the apparent treatment effect will shrink, and comparing therapies will get harder. Read the full post: na2.hubs.ly/H08v_0d0 #Obesity #ClinicalTrials #DigitalTwins #ClinicalDevelopment

🎉 We’re excited to welcome Ian Fleming as Unlearn’s new VP of Product. Ian brings 25 years of experience in the life sciences industry. He started with hands-on statistical work on clinical programs, then moved into building the technology that clinical and biometrics teams rely on, most recently building a team that delivered foundational components of a major clinical data platform and its AI infrastructure. Welcome to the team, Ian!

Sponsors are no longer asking whether digital twins work. They’re asking who builds the infrastructure, what data it runs on, and whether it will hold up when regulators ask about it years from now. Many internal builds stall on the same three things: single-study models that can’t follow a program as it evolves, the cross-functional alignment needed to get twins into a protocol, and harmonized data that everything else depends on. In our latest post, we share what we’ve learned from working with leading sponsors, including why the right unit for a model is the disease, not the study. If your team is building this now, or has tried and stalled, we’d like to compare notes. na2.hubs.ly/H08rdMf0 #DigitalTwins #ClinicalDevelopment #AIinClinicalTrials

📄 Our latest whitepaper, A Field Guide to Digital Twins in Clinical Development, walks through each decision point in a study, from eligibility criteria and sample size in design, to stopping rules at interim, to interpreting the result after readout, and shows how digital twins fit at every step. It also covers the methods behind each use and the deployment principles we follow, drawn from the @FDA's draft guidance on AI in drug development. Download the field guide: na2.hubs.ly/H08nFyK0 #ClinicalTrials #DigitalTwins #DrugDevelopment #Biostatistics #AI

Six years of working with @FDA and @EMA_News on #digitaltwins for #clinicaltrials, summarized. Luis Olmos on where the #regulatory work stands and what it means for your next study: na2.hubs.ly/H08gQnt0

One of the biggest problems in #obesity #clinicaltrials today is that placebo patients can figure out their assignment quickly and take action into their own hands if they’re not losing weight. They can leave the study, or start treatment on their own…all while their results still count toward the placebo arm. Our co-founder and Chief Scientific Officer, Jon Walsh, wrote about what that does to the data and why the fix belongs in #trialdesign. Read the post: na2.hubs.ly/H08ftx20

🇵🇱 Heading to #EHDN2026 in Kraków, October 22–24? Catch our presentation, "AI-Powered Digital Twins: Applications and Opportunities in Clinical Trials." Huntington's disease progresses differently in every patient, which makes treatment effects hard to detect. A #digitaltwin for each participant helps: more power, a smaller control arm, and more participants on the investigational treatment. In planning one HD registrational study, digital twins reduced the required sample size by up to 29%. Connect with us in Kraków.

More drug candidates than the trial system can absorb. Then what? The takeaways from our panel with Tala Fakhouri (Parexel, formerly FDA), Barbara Bierer (Harvard Medical School, MRCT Center), and Unlearn’s Co-founder and CSO, Jon Walsh, plus the recording: na2.hubs.ly/H089S910

💊 Every clinical trial is a chain of decisions, and you don't have to wait for the next protocol to start making them with better evidence. In a new post, our Co-founder and Chief Scientific Officer Jon Walsh lays out the three points in a trial where sponsors can come in. He also names a mistake he often sees teams make when building AI for drug development. The principles behind machine learning were never written down, so he wrote down three of them, along with why they matter when the data is private and the stakes are a go/no-go decision. This post also opens a new chapter in how Unlearn works with sponsors: data, digital twins, and AI working together at all three points. Read Jon's post: na2.hubs.ly/H089c_20 #ClinicalTrials #DigitalTwins #DrugDevelopment #AI #MachineLearning