Joint efforts kick off with DREAM Challenge that allows scientists to compute on NIH’s Covid data in a privacy-enhancing environment and experience an unprecedented level of transparency in generative AI
BeeKeeperAI., a pioneer in privacy-enhancing, multi-party collaboration software for AI development and deployment, and cStructure, a leading innovator in collaborative causal inference, announced a collaboration for advancing causal AI to speed up the ability of scientists, health data stewards and AI algorithm developers to build and train AI models that can be trusted for scientific advancement and healthcare innovation.
Health Technology Insights: Stylus Medicine Launches to Advance Precision In Vivo Genetic Medicines
Causal AI is one of the next big frontiers for GenAI to be successful and more widely trusted in the healthcare and scientific fields − going far beyond just identifying patterns as traditional AI is known to do.
The two companies are pioneering a novel, causal AI-centric approach that preserves patient data privacy, while enabling transparency, at scale, to better understand and model cause-and-effect relationships within health-related data, including data from large populations. The rigorous data analysis is captured in causal graphs that can be reliably used in high-quality, regulatory-grade life science. Ultimately, causal AI makes GenAI more trustworthy and compliant with regulatory-based best practices.
To kick off this collaboration, BeeKeeperAI and cStructure are launching the “Covid Causal Diagram DREAM Challenge,” a crowd-sourcing initiative that opens up access for scientists to analyze real-world COVID data from the National Institutes of Health (NIH) in a privacy-enhancing environment for the purpose of accelerating the determination of the causal relationships between treatment and patient outcomes.
Health Technology Insights: Castlight Health Expands Partnership with ArcBest for Primary Care
Causal AI is one of the next big frontiers for GenAI to be successful and more widely trusted in the healthcare and scientific fields − going far beyond just identifying patterns as traditional AI is known to do. Causal AI is an answer to the struggles of GenAI to consistently deliver the best possible factual information. Causal AI is designed to explain why something happened and what will happen. It is also a key to speeding up how regulatory bodies, such as the FDA, evaluate clinical studies that use AI. Critical to the advancement of causal AI are transparency, data privacy, efficiency and global access to real-world data.
“Our collaboration with cStructure is a perfect match to leverage GenAI for innovation at the speed of industry, combining BeeKeeperAI’s privacy-preserving EscrowAI data platform and cStructure’s causal diagram tech to accelerate the adoption of causal AI for improving human health,” said Dr. Michael Blum, MD, Co-founder and Chief Executive Officer at BeeKeeperAI. “We’re excited about our first project with cStructure to address the challenges of AI in health-centric applications. Through the DREAM Challenge, biomedical scientists will be able to compute on NIH data in our privacy-enhancing EscrowAI environment and have teams collaborate to build causal graphs. The results have the potential to change the way that the community thinks about causal relationships and transparency of AI in healthcare.”
Erick R. Scott, MD, Founder of cStructure, said, “We have made significant progress at cStructure in establishing a collaborative interface for developing causal diagrams that visually represent treatment effects, confounders, and outcomes. A necessary complement is a secure collaboration environment where AI models can compute on sensitive data while preserving privacy and protecting the intellectual property of the model. BeeKeeperAI delivers a privacy-preserving platform that automates the use of confidential computing, which provides the highest level of security for AI. We’re proud to partner with BeeKeeperAI on the Covid Causal Diagram DREAM Challenge and on the opportunity to make causal AI mainstream for science and healthcare.”
The DREAM Challenge
Covid Causal Diagram DREAM Challenge asks participants to develop Structural Causal Models (SCMs) with the assistance of Large Language Models (LLMs). This causal diagram challenge focuses on the effect of glucocorticoids on 28-day survival rates in hospitalized COVID-19 patients. The challenge opens May 15, 2025.
The cStructure platform provides a collaborative causal graph interface that participants will use to develop models specifying relationships between patient characteristics, treatments, and outcomes. An integrated large language model assistant offers domain expertise and support.
Challenge participants will have access to real-world data collected in NYC during the early stages of the global pandemic to train their SCMs. At the end of the challenge, submitted models will be securely evaluated within the remote EscrowAI enclave using data collected during a fit-for-purpose cohort study to simulate federated learning.
“We at DREAM Challenges are thrilled to work with cStructure and BeeKeeperAI on the Covid Causal Diagram DREAM Challenge — a groundbreaking effort and bold step toward harnessing causal AI to answer critical questions in medicine. By combining privacy-preserving technology with global scientific collaboration, we are advancing a future where AI not only predicts but explains, driving real breakthroughs for patient care,” said Gustavo Stolovitzky, PhD, Founder, Chair Emeritus and Director of DREAM Challenges.
Models will be assessed on the following criteria: comparison with high-quality randomized controlled trial (RCT) results; proper adjustment for confounding; and plausibility of causal relationships in the submitted model.
Health Technology Insights: Medable Launches Digital Oncology Trial Solution to Cut Complexity
To participate in our interviews, please write to our HealthTech Media Room at sudipto@intentamplify.com
Source – businesswire
