Most collaboration starts with sharing. In biomedical research, however, valuable information—such as genetic data and clinical records—is often locked within the institutions that collected them, protected by strict patient privacy policies. While access can be negotiated, the administrative friction involved in pooling data from multiple institutions can create a major bottleneck, potentially stalling urgent discoveries.
“If researchers can analyse health data without moving or exposing underlying raw records, they can remove the biggest source of delays in bioinformatics, and make collaborations happen that simply wouldn’t happen otherwise,” said Weizhuang Zhou, Senior Assistant Director at A*STAR’s Scientific Data Strategy Office (SDSO).
To realise that possibility, researchers at the A*STAR Institute of Advanced Intelligence and Computing (A*STAR IAIC) are building a new federated health data analytics platform, working in collaboration with the A*STAR Genome Institute of Singapore (A*STAR GIS), A*STAR Bioinformatics Institute (A*STAR BII), National Heart Centre Singapore, National University of Singapore (NUS) and Duke-NUS Medical School.
The team’s platform runs statistical tests across encrypted data held by different institutions. To do so, the system combines two cryptographic tools: multi-party homomorphic encryption (MHE), which handles the heavy statistical calculations that dominates most analyses; and secure multi-party computation (MPC), which steps in for trickier operations such as division.
“Imagine a group of friends who want to know their average salary without revealing what they earn,” said Chao Jin, a Principal Scientist at A*STAR IAIC, which consolidates the former A*STAR Institute of Infocomm Research (A*STAR I2R) and A*STAR Institute of High Performance Computing (A*STAR IHPC).
Ordinary encryption would keep each salary private, but make the average impossible to calculate. The team’s approach runs calculations directly on the locked data, revealing only the final combined statistic to all parties.
“Through this approach, no hospital sees another's raw numbers, no individual patient's data is exposed, and no central copy of everyone's data is ever created,” Jin added.
Zhou, Jin and colleagues tested their platform on two Singapore cohorts—SingHEART and ATTRaCT—running encrypted ANOVA tests, t-tests and meta-analyses that returned results identical to standard unencrypted analysis. In separate benchmark tests on a scale comparable to the UK Biobank, the team’s approach completed a secure ANOVA test in under two minutes, outperforming other platforms built on a pure MPC-based approach.
The team then applied their platform on a real-world biomedical question, evaluating three polygenic risk scores for hypertension that were developed mainly from European populations.
They found that two scores—PGS000957 and PGS000958—tracked measured blood pressure more clearly in Singaporean cohorts than the third, echoing prior findings that such scores often generalise poorly beyond the populations they were built on.
The team’s next step is deploying their approach on Singapore’s Biomedical Data Hub to enable secure cross-institutional studies. “This extends the analyses that platforms like BDH can safely support and reduces the reliance on slow, case-by-case data-transfer agreements that currently gate such studies,” Jin added.
The A*STAR-affiliated researchers contributing to this research are from the A*STAR Institute of Advanced Intelligence and Computing (A*STAR IAIC), A*STAR Genome Institute of Singapore (A*STAR GIS) and A*STAR Bioinformatics Institute (A*STAR BII).
