A public-interest, open-source project
Cancer care,
in context.
Reports, treatment histories and molecular knowledge belong in a clearer picture. MyOncode helps patients and caregivers keep care information connected, with tools for users to check and correct AI-extracted information.
The project
A record that follows the course of care.
MyOncode (知见) brings reports, treatment history, symptoms and follow-up information into a longitudinal record. Patients and caregivers check AI-extracted information against their own documents, correct entries, revisit the timeline and prepare for conversations with their care teams.
Alongside the record workflow, we are developing reference content on gene variants, molecular targets and signaling pathways to make molecular context easier to explore and understand. Totoro reviews this educational content.
Research
Extraction quality and user verification.
Our research brings together bioinformatics, molecular oncology knowledge representation and evaluation of AI-assisted records. We study how molecular context can accompany longitudinal oncology information, and how users check and correct AI-extracted records. The evaluation design uses fully synthetic records and reference annotations, so the work can be inspected and reproduced.
- Extraction quality
- Field-level precision and recall, unsupported additions, and assignment to the correct treatment stage.
- User verification
- The corrections users make to AI-extracted records, remaining extraction errors, and time spent checking the information.
- Molecular context
- Reference content connecting gene variants, molecular targets and signaling pathways with oncology information.
Open research outputs
The work is organized around a public evaluation protocol, an annotated synthetic benchmark, reproducible scripts and open-source technical reports.
Get in touch
Totoro — Founder & Research Lead, MyOncode
Totoro leads MyOncode's work in bioinformatics and molecular oncology knowledge representation, and reviews its educational content on gene variants, molecular targets and signaling pathways.
Selected for OpenAI's Codex for Open Source program. Get in touch about research collaboration or contributing to the project.
myoncode@0xtotoro.com