Phenotypic Similarity-based Approach For Variant Prioritization For Unsolved Rare Disease: A Preliminary Methodological Report

Solve-RD is a European-funded research project that aims to molecularly solve unsolved cases of rare diseases (RD) by using different data and re-analysis approaches. The project builds upon four European Reference Networks (ERNs) which annually see more than 270,000 RD patients, and contribute unsolved patients data. This article describes a methodology based on phenotypic similarity calculations among solved/unsolved cases and known RD, using phenotypic annotation comparisons, Natural Language Processing, Artificial Intelligence, Machine Learning, Deep Neural Networks and entity recognition in clinical narratives. This led to the identification of a formerly undescribed disease, or to the identification of unreported manifestations of known RD, with the goal of returning a clinical diagnosis to the patients.