Case of Miriam González: When AI Meets Ultra-Rare Cancer, Exploring the Limits of Personalized Medicine

This article introduces the ultra-rare cancer case of Miriam González: metastatic breast cancer with 80% neuroendocrine differentiation and 13-fold FGFR1 gene amplification, which is unresponsive to standard treatments. The Beyond The Protocol open-source project, in collaboration with an international team, built a medical record tracking platform through AI assistance and global collaboration to explore breakthrough therapies. The project uses the Nuxt 4 tech stack, emphasizes medical transparency and community participation, and provides a new paradigm for rare disease treatment.

Miriam González Pérez is a metastatic breast cancer patient whose case has two extremely rare molecular features: approximately 80% neuroendocrine differentiation (fast tumor growth, poor prognosis, and poor response to conventional chemotherapy/endocrine therapy) and 13-fold FGFR1 gene amplification (drives tumor progression, and related treatments are still in the exploratory stage). This combination is extremely rare in breast cancer, with no efficacy data for standard protocols, posing huge challenges to treatment.

Faced with the limitations of standard treatments, Miriam's case gave birth to the Beyond The Protocol open-source project. Its core concept is to use AI and global collaboration to break through treatment bottlenecks. The team consists of international medical experts, data scientists, and developers, building an information tracking platform that integrates medical record data, treatment history, and research progress, serving as an open collaboration space to invite the global medical community to participate in the exploration of rare cancer treatments.

The project uses the Nuxt4 framework, with Nuxt Content v3 for content management; built-in @nuxtjs/i18n supports Spanish and English bilingualism, following accessibility design; integrates SEO optimization and the nuxt-ai-ready module; synchronizes GoFundMe fundraising data in real-time via Netlify Functions (CDN cache for 15 minutes/1 hour), balancing information freshness and API call efficiency.

AI assistance value: quickly screening literature to identify relevant treatment plans, analyzing similar cases to find correlations, and integrating global research progress for decision support. Real-world challenges: scarcity of training data for rare subtypes limits supervised learning effects, correlations found by AI need clinical validation, and AI recommendations cannot replace doctors' professional judgment.

The project is fully open-source; GitHub public code and content break information barriers; community collaboration mechanisms (Issues/Pull Request) promote continuous improvement; replicability allows other rare disease patients/organizations to fork and build platforms, empowering the rare disease community.

Miriam's case triggers thinking about the dilemma of rare disease treatment; the project fills treatment gaps through information integration and collaboration; real-time fundraising data solves crowdfunding trust issues; AI is positioned as an auxiliary tool, and collaboration with human doctors is the optimal path. Conclusion: The project demonstrates a transparent, open, patient-centered medical information model, inspiring more people to participate in rare disease exploration.