About LYFE Sciences

Built by a genome analyst, for everyone

LYFE Sciences started as one analyst's attempt to stop opening eight database tabs for every variant. It is still that, with a pipeline behind it.

LYFE Sciences Project: HERA Since 2023
01

How it started

LYFE Sciences began as a personal project in 2023 to simplify genetic variant interpretation in day-to-day work. As a genome analyst, I frequently encountered the challenge of quickly accessing comprehensive and reliable genetic variant information.

To address that, I built LYFE Sciences: an accessible, user-friendly tool that consolidates the crucial genetic data in one place.

02

What I'm trying to do

My main goal is to streamline variant analysis by reducing the need to navigate between multiple resources or pay for third-party applications.

I continually update LYFE Sciences with the latest AI architectures and enhancements to keep it practical, relevant, and useful for everyday clinical and research work.

03

How it works

LYFE Sciences is designed around the real workflows of genomic professionals. Rather than requiring users to juggle ClinVar, OncoKB, JAX-CKB, COSMIC and other databases independently, Project HERA aggregates, normalises and presents that data in a unified interface.

Every query runs against up-to-date sources, so interpretations reflect the latest evidence rather than a static snapshot from months ago.

The tool is continuously refined based on real-world usage and the evolving landscape of genomic databases.
Try a variant query
04

The next step: Project HERA

Project HERA is the third iteration of LYFE Sciences: a modular AI pipeline for variant interpretation that integrates retrieval, normalisation, framework selection, criterion assessment and structured result generation into a single workflow.

It is designed around the idea that reliable AI needs controlled inputs, explicit evidence mapping, framework-aware decision logic, and outputs that are easy to audit and operationalise.

The pipeline combines automated evidence gathering with agentic decision making, prioritises ClinGen and gene-specific specifications, and produces structured review artifacts for downstream systems.

01
Retrieval
02
Normalisation
03
Framework selection
04
Criterion assessment
05
Structured output

Project HERA is an example of how AI-native pipelines can bridge language models, domain-specific logic and production software architecture without sacrificing the traceability that clinical use demands.

HERA is an AI system, and it can make mistakes. Traceability is the point of the architecture above — every criterion carries its sources so a classification can be checked rather than taken. Review the evidence before relying on any call.
05

Why it matters

By sharing LYFE Sciences, I hope to make genetic analysis less overwhelming, and help others like myself improve their work.

Faster variant interpretation means analysts spend less time hunting for information and more time on the judgment the job actually needs.

matthewlyf.com