Case Study: Miroculus accelerates microRNA biomarker discovery for gastric cancer with Neo4j

Machine Learning and Graph Technology Accelerate Medical Research

Miroculus, a San Francisco–based molecular diagnostics company, set out to use circulating microRNAs to create a less invasive blood test for gastric cancer but faced a huge barrier: an exploding body of medical literature and complex, time-consuming methods for identifying relevant microRNA–gene–disease relationships. Early detection of stomach cancer is critical, yet current screening (like endoscopy) is invasive and inefficient, so Miroculus needed a faster way to connect scientists to the latest research and find reliable biomarkers.

To solve this, Miroculus ingested over a billion articles into Hadoop, used NLP to extract candidate sentences, and applied an unsupervised machine learning model to infer relationships that were stored in a Neo4j knowledge graph and exposed via an interactive visualization. The pipeline accelerated insight discovery, supported an FDA-guided study with NIH/NCI collaborators (650 participants), and led to identification of microRNAs that screen for gastric cancer from blood—offering a minimally invasive alternative and publicly available tools for researchers.

Miroculus

Antonio Molins

VP of Data Science