DNA Methylation Pipeline

For a lot of rare-disease families, the hardest sentence to hear is "the sequencing came back inconclusive." You've run the genetic test that was supposed to end the search, and it found nothing nameable. So where do you look next?

DNA methylation is one good answer. Think of your genome as a long text: the letters are the DNA sequence, but the body also writes small chemical notes in the margins that change how a passage gets read, without changing the letters themselves. Those notes are methylation marks, and certain genetic diseases leave a distinctive pattern of them across the epigenome. Read the pattern and you can sometimes name a condition that years of standard testing couldn't.

This pipeline reads those patterns. I built it from both sides of the problem: as an ML scientist with a background in bioinformatics and computational biology, and as a rare disease patient who spent roughly five years undiagnosed, watching a healthcare system treat me like a problem it wasn't built to solve.

A working clinical pipeline isn't a demo. It's infrastructure, and infrastructure for diagnosis touches patient outcomes directly. This one connects to the Stanford RTTP collaboration, the Mayo Clinic Undiagnosed Patients Hackathon, and the wider Wilhelm Foundation rare disease AI ecosystem, all of them working the same seam: the gap between what the technology can already do and what actually reaches the person in the exam room. Behind every methylation signature in the dataset is someone still waiting for their answer.

I built the tool that might have diagnosed me. That's not a marketing line. That's the structural truth of this project, and it's the reason I keep building it.