RINAE.AI
A workbench for the diseases too rare for anyone else to build for.
Identify the target. Score the odds. Design candidate molecules.
Example workspace — illustrative, not a real analysis. The findings below are sample data, not a RINAE-produced result.
Argus ◉ pulling evidence from 15 bioinformatics sources — ClinVar, gnomAD, GTEx, OMIM…
The Problem
Ultra-rare disease isn't just hard. It's economically broken.
Conventional pharma is built around indications with a million patients and billion-dollar revenue ceilings. None of that math applies when fewer than a thousand children share the same diagnosis worldwide.
<1,000
Patients globally for ultra-rare
An eligible trial pool measured in dozens, not thousands.
5–7 yrs
Symptom onset to genetic diagnosis
8+ specialists consulted before the right answer arrives.
3–5 yrs
To enroll a 30-patient trial
KOL networks, conferences, and advocacy chains identify ~5 patients/month.
$50–150M
Traditional per-program cost
Down from $1–2B for common indications, but still unviable at this scale.
The brutal math. At 500 patients globally with a 50% diagnosis rate and 70% treatment eligibility, the addressable population for a single ultra-rare program is roughly 175 patients. Traditional development costs of $100M+ make that marginal at best — which is why most of these diseases never get a program.
The Approach
Why traditional approaches fail.
For ultra-rare disease, the steps that already take years for a common indication break down entirely. Here is where the conventional model stalls.
Patient finding
The rare-disease diagnostic odyssey averages years.
TraditionalKOL networks, conferences, advocacy chains identify ~5 patients/month.
Natural history
Most ultra-rare diseases have no documented natural history at all.
TraditionalProspective studies with limited sites and selection bias.
The Pipeline
Three tools. One surface.
Each stage is a real service the team ships today. The platform wires them into a single pipeline with shared state, review gates, and an audit trail.
Stage 1Argus
Pull evidence.
Query 15+ bioinformatics databases for a gene target — expression, variants, constraint, pathways — and materialize one normalized evidence packet.
Stage 2Admiral
Assess feasibility.
Multi-agent amenability scoring: five specialist models deliberate on whether the target is druggable with an ASO and produce a structured report.
Stage 3Metamorph ASO
Design ASOs.
Generate 20-nt antisense oligonucleotide candidates, fold and rank them, and screen off-target binding against the human transcriptome.
“My daughter Rose was diagnosed with HNRNPH2-NDD when she was three. We started RINAE because Rose couldn’t wait for the industry to care.”
Casey McPherson
Founder · AlphaRose Therapeutics
HNRNPH2 — Lead Program
Published HNRNPH2 epidemiology estimates — disease-landscape context, not RINAE-produced figures.
By the traditional route, enrolling a 30-patient trial can take 3–5 years and exhaust a program's runway before generating meaningful data.
Who built this
Genzyme, Alnylam, Krystal Biotech alumni.
The team that built Vyjuvek — the first FDA-approved in-vivo gene therapy — is operating the platform alongside rare-disease researchers and AI engineers.
Alan Walts
Executive Chairman
27 years at Genzyme · President, Genzyme Pharmaceuticals
Belinda Termeer
Co-founder
Termeer Foundation · Genzyme alumni network
John Garcia
CCO
VP at Alnylam · SVP at Krystal Biotech (launched Vyjuvek)
Reference corpus
188K ASO patents
ASO Atlas: 417 distinct chemical designs distilled into Admiral's reasoning.
Why we exist
Built to make ultra-rare programs economically possible — the ones the traditional model writes off.
One workflow
From genetic signal to a designed candidate — one connected workflow instead of fragmented, vendor-by-vendor handoffs.