A Secure Polygenic Risk Score (sPRS) is a privacy-preserving protocol that computes an individual's genetic susceptibility to a specific disease by securely combining their private genomic data with a proprietary model of aggregated variant effects, ensuring that neither party reveals their sensitive input to the other. This is achieved through secure multi-party computation (MPC) techniques, such as secret sharing and garbled circuits, which allow the calculation to be performed on encrypted data.
Glossary
Secure Polygenic Risk Score
A privacy-preserving protocol for calculating an individual's genetic susceptibility to a disease by securely combining their genomic data with a proprietary model of aggregated variant effects using secure multi-party computation.
PRIVACY-PRESERVING GENOMICS
What is Secure Polygenic Risk Score?
A cryptographic protocol for calculating disease susceptibility from genetic data without exposing the individual's genome or the proprietary risk model.
The protocol typically involves a patient or research participant holding an encrypted genotype and a model owner holding a confidential weight vector for thousands of single nucleotide polymorphisms (SNPs). Using cryptographic primitives like Beaver triples for efficient multiplication and DReLU protocols for non-linear activation, the system computes a weighted sum without ever decrypting the raw genome or exposing the model's proprietary coefficients, enabling collaborative genomic research under strict data governance.
ARCHITECTURAL COMPONENTS
Key Features of Secure PRS
A Secure Polygenic Risk Score protocol decomposes into distinct cryptographic and bioinformatic modules that collectively ensure both input privacy and computational integrity.
01
Secret-Shared Genotype Input
The individual's genomic variants are additively secret-shared across two or more non-colluding computing servers before any computation begins. No single server ever sees the raw genotype.
- Uses Shamir's Secret Sharing or simple additive replication
- Input sharing occurs client-side in a trusted local environment
- Prevents the model owner and compute parties from learning the individual's genetic predispositions
Zero
Raw Genotype Exposure
02
Proprietary Effect-Weight Model
The aggregated variant effect sizes—the intellectual property of a research institution—are also secret-shared among the computing parties. The model remains encrypted throughout the entire calculation.
- Protects the GWAS-derived odds ratios and linkage disequilibrium matrices
- Model weights are never reconstructed in the clear
- Enables commercial monetization of proprietary risk models without intellectual property leakage
03
Secure Linear Combination
The core computation is a privacy-preserving dot product between the secret-shared genotype vector and the secret-shared effect-weight vector, executed using Beaver triples for efficient multiplication.
- Leverages pre-computed multiplication triples to eliminate online cryptographic overhead
- Implements secure truncation to manage fixed-point precision
- Population-level normalization factors (means, standard deviations) are applied obliviously
04
Oblivious Risk Stratification
The final PRS is compared against population percentile thresholds using a secure comparison protocol (DReLU), outputting only a risk category to the authorized recipient without revealing the raw score.
- Prevents gradient leakage or inference of the underlying model
- Output can be a simple categorical result (e.g., 'elevated risk')
- The individual learns their risk tier but nothing about the proprietary model's internal thresholds
05
Auditable MPC Ceremony
The entire protocol execution generates a cryptographic transcript that can be verified by a third-party auditor without revealing the private inputs, ensuring compliance with regulatory frameworks.
- Implements verifiable secret sharing for input consistency checks
- Provides non-repudiation for clinical decision support contexts
- Supports integration with sovereign health data infrastructure requirements
06
Population Reference Panel Security
Ancestry-specific allele frequency and linkage disequilibrium reference panels, required for score normalization, are accessed via Private Information Retrieval (PIR) to hide which population stratum is being queried.
- Prevents the reference panel server from learning the individual's genetic ancestry
- Uses keyword PIR or index PIR for efficient oblivious lookups
- Mitigates side-channel leakage through population-specific reference data access patterns
PRIVACY-PRESERVING GENOMICS
Frequently Asked Questions
Clear, technically precise answers to the most common questions about secure polygenic risk score computation, a critical intersection of cryptography and precision medicine.
A Secure Polygenic Risk Score (sPRS) is a privacy-preserving protocol that calculates an individual's genetic susceptibility to a specific disease by securely combining their private genomic data with a proprietary, aggregated model of variant effects, without revealing the raw genome to the model owner or the model weights to the individual. The computation typically leverages secure multi-party computation (MPC) or homomorphic encryption. In an MPC-based architecture, the individual's genotype and the model owner's effect-size weights are secret-shared among two or more non-colluding computing servers. These servers jointly perform a linear weighted sum—multiplying each single nucleotide polymorphism (SNP) by its corresponding effect size and summing the results—over the secret-shared values. The final risk score is reconstructed only by the authorized recipient, usually the individual or their physician. This ensures that the model owner learns nothing about the individual's genetic variants, and the individual learns nothing about the proprietary model coefficients beyond their final aggregated score.
Enabling Efficiency, Speed & Accuracy
Intelligent Analysis, Decision & Execution
We build AI systems for teams that need search across company data, workflow automation across tools, or AI features inside products and internal software.
Talk to Us
Search across company data
Give teams answers from docs, tickets, runbooks, and product data with sources and permissions.
Useful when people spend too long searching or get different answers from different systems.
Enterprise searchRAGPermissions
Read more
Automate internal workflows
Use AI to route work, draft outputs, trigger actions, and keep approvals and logs in place.
Useful when repetitive work moves across multiple tools and teams.
AI agentsWorkflow automationGovernance
Read more
Add AI to products and internal tools
Build assistants, guided actions, or decision support into the software your team or customers already use.
Useful when AI needs to be part of the product, not a separate tool.
AI integrationDecision supportModel routing
Read more
About the author
Prasad Kumkar
CEO & MD, Inference Systems
Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.
His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.
LinkedIn
Limited slotsGet a Free AI Consultation
Partnered with leading AI, data, and software stack.
How We Work
Custom AI workflows for your Business
One-fit-all AI don't work for modern businesses. At Inferensys, we aim to understand your business & custom requirements; which we use to define most efficient agentic workflows, the data, and the tools for your business.
01
Review the use case
We understand the task, the users, and where AI can actually help.
Read more02
Pick the right approach
We define what needs search, automation, or product integration.
Read more03
Build the first useful version
We implement the part that proves the value first.
Read more04
Improve from there
We add the checks and visibility needed to keep it useful.
Read moreThe first call is a practical review of your use case and the right next step.
Talk to Us