An eight-initiative framework for AI-accelerated rare disease research.
243 of 245 pathogenic WFS1 missense variants — 99.2% — leave the wolframin fold largely intact.
“Most pathogenic missense mutations in WFS1 do not break the wolframin protein. They break specific contacts on a protein that still folds.”
Battle-tested with leading Wolfram syndrome researchers and designed to generalize across any rare disease. Each initiative draws on infrastructure that already exists — assembled, not invented. From immediate deployments to long-horizon breakthroughs.
Turn the six-year Wolfram diagnostic odyssey into a flagged alert at first presentation — connecting an AI rare disease diagnostic system (DeepRare) into hospital EHR infrastructure (Epic) to pattern-match symptoms across organ systems.
The first complete open structural atlas of pathogenic WFS1 missense variants — 245 variants mapped in 3D on Google DeepMind's protein-structure AI, clinically annotated. 99.2% are fold-intact and likely small-molecule rescuable.
Turn the WFS1 atlas into a treatment-matching engine. A graph neural network trained on 17,000 diseases screens approved compounds against every variant. Sarah Gladstone and Nufar's first 25–30 candidates — now in zebrafish testing — are the human-validated foundation the AI extends.
Deploy a federated learning platform built for biopharma scale (NVIDIA FLARE) to connect 5,000 Wolfram patients across 6 countries — letting the model cross borders without a single medical record ever leaving its home institution.
Pair candidate biomarker panels with a retinal AI trained on 1.6 million images (RETFound) to find molecular signals that replace clinical decline as the trial endpoint — compressing 18-month readouts into months.
An LLM-powered Wolfram research engine — every paper, preprint, and trial update curated from PubMed and beyond, with document upload so clinicians can query their own research against the corpus. On May 29, Dr. Sarah Gladstone uploaded an atlas PDF and surfaced three therapeutic strategies — one already in active testing.
Architecting Wolfram trials with AI-powered synthetic control arms and Bayesian adaptive design. The FDA already accepted this approach with AMX0035 — making approvable Wolfram trials possible at rare disease scale.
Combine Wolfram patient stem-cell lines (iPSCs), NIH NCATS pharmacological-grade robotic imaging (Cell Painting), and the WFS1 Molecular Atlas into one structure-guided drug screen — the most targeted compound test Wolfram has ever had.
Every known pathogenic WFS1 missense variant in ClinVar — 245 mutations — with an AlphaFold structure, AlphaMissense pathogenicity score, ClinVar evidence, DynaMut2 stability prediction, an interactive 3D viewer, and an editorial narrative. The world's first complete open structural atlas of a pathogenic monogenic disease landscape.
Wolfram Research Intelligence — an LLM-powered research engine trained on the full Wolfram literature. Query the corpus, upload your own papers and data, and get synthesis in seconds. Delivering an automated weekly research digest to Snow Foundation leadership since April 2026.
Mission-driven foundations with urgent research goals but limited in-house AI capability. We are your infrastructure team.
Fragmented global researcher networks — we build the shared intelligence layer that connects what you know.
Companies pursuing orphan drug designations who need AI-accelerated target discovery and trial design.
NIH, WashU, Cambridge and peer institutions conducting rare disease research who need scalable AI support.
RareResearch.AI was founded by Mike Wallace — entrepreneur, culture architect, and AI strategist. Born from a live framework presentation to leading Wolfram syndrome researchers at The Snow Foundation, the company is the product of a room full of scientists saying: "This model applies to hundreds of rare diseases."
Mike bridges the gap between complex AI infrastructure and the rare disease communities that need it most — operating at the intersection of deep science and human story.
“This model, Mike, you can apply these eight categories to a huge number of rare diseases. I think you just found your new company.”
Whether you're a rare disease foundation, a researcher, or an investor — every community that engages with this framework makes the model stronger for all of them.