SelfDecode set out to redefine consumer genetics by delivering ancestry-adjusted polygenic risk scoring (PRS) — translating raw DNA data into actionable, multi-variant health recommendations rather than single-gene panic reports. Elementica engineered the backend data pipelines connecting raw genotype file ingestion, automated SNP imputation, and dynamic report generation, enabling the platform to process consumer files from 23andMe, AncestryDNA, and MyHeritage into trusted insights now used by major healthcare providers, including reported use at Cleveland Clinic.
Healthcare/AI/Genetics/Precision Medicine
DTC interoperability — seamless ingestion for 23andMe, AncestryDNA, and MyHeritage raw files
Enterprise adoption by major institutions, including reported use at Cleveland Clinic
Zero-panic UX — polygenic netting engine balancing protective vs. risk variants to prevent false alerts
Client context
Early consumer genetics platforms suffered from a fundamental flaw: "one gene, one disease" reporting. Flagging a single isolated single-nucleotide polymorphism (SNP) often triggers unnecessary patient panic while ignoring the broader genomic context. SelfDecode pioneered a scientifically rigorous approach using Polygenic Risk Scoring (PRS), aggregating signals across thousands to millions of genetic variants while adjusting for a user's ancestral background to produce accurate, personalized health recommendations.
The challenge
Converting cutting-edge genomic science into a scalable consumer SaaS platform presented two distinct engineering barriers:
- Heterogeneous & Incomplete Raw Data: Consumer genetics providers (23andMe, AncestryDNA, MyHeritage) use different genotyping microarrays, covering varying subsets of the genome (~700,000 SNPs out of tens of millions). The engine needed to normalize and expand these disparate file formats reliably.
- Complex Data Ingestion at Scale: Processing multi-megabyte raw genotype text files, executing heavy statistical calculations, and generating prioritized, readable health recommendations required significant backend compute without causing user-interface timeouts or soaring infrastructure costs.
Elementica's role
Elementica led the platform engineering that connects raw user file uploads to the underlying scientific scoring models. We built the high-throughput ingestion architecture, integrated the genotype imputation pipeline, and engineered the rendering layer that transforms raw scientific outputs into prioritized health action plans.
The solution
Universal DNA Ingestion & Imputation Pipeline
We built a resilient file-ingestion system capable of parsing and validating raw text files from any major direct-to-consumer genetics provider. Once uploaded, files are routed into an automated imputation pipeline. Using reference panel populations, the engine statistically infers millions of unassayed SNPs, expanding a ~700K SNP file into a rich dataset required by SelfDecode's polygenic scoring algorithms.
Variant Netting & Ancestry-Adjusted Scoring Engine
Behind the reporting interface, we implemented logic that weighs positive, protective genetic markers against risk-conferring variants simultaneously. The engine calculates an ancestry-adjusted baseline score, ensuring that individuals of diverse ethnic backgrounds receive accurate risk calculations rather than false positives caused by unrepresentative European-centric genomic reference sets.
Dual-Surface Clinical & Consumer Reporting
We engineered a flexible frontend delivery layer that dynamically adapts statistical confidence intervals based on the user type. Consumer profiles display prioritized, intuitive lifestyle and dietary recommendations, while the practitioner portal (PromicsEdge) surfaces granular variant tables, odds ratios, and scientific references tailored for clinical decision-making.
Engineering decisions that mattered
Decoupled asynchronous imputation via cloud batch queues
Running genomic imputation on web application servers would cause severe latency spikes and memory exhaustion. We decoupled the imputation workflow using AWS Batch queues triggered by S3 upload events. This serverless batch architecture automatically scales compute nodes based on incoming upload volume, keeping API web servers responsive and infrastructure spend optimized.
Imputation before scoring to unlock universal data compatibility
Rather than forcing users to purchase expensive whole-genome sequencing (WGS), we prioritized building a robust imputation pipeline upstream of the scoring engine. Investing in high-confidence statistical imputation allowed SelfDecode to accept low-cost array data from any market vendor while maintaining clinical-grade analytical precision.
Results
SelfDecode established itself as a precision health platform, delivering science-backed risk reports across more than 1,500 health conditions. The underlying infrastructure successfully powers both consumer-facing subscriptions and the enterprise-grade PromicsEdge practitioner tool, supporting adoption by leading healthcare organizations like the Cleveland Clinic.
Product evolution
The underlying imputation and scoring engine built during this project continues to expand, supporting whole-genome sequencing (WGS) ingestion, real-time lab test integration, and automated AI-driven diet and supplement planning.
