RDmaster

RDmaster performs phenotype-oriented differential diagnosis of rare diseases by integrating two Bayesian diagnostic methods and adaptive phenotype interrogation to generate prioritized candidate diagnoses.

Key Features:

• Bayesian Diagnostic Methods Integration: Incorporates two Bayesian diagnostic methods to generate a candidate list of potential diagnoses and to improve interpretability for differential diagnosis (DDX).
• Adaptive Information Gain and Gini Index (AIGGI): Implements AIGGI to evaluate expected information gain from interrogated phenotypes in real time and prioritize diagnostically informative features.
• Question-and-Answer (Q&A) Dialogue Strategy: Utilizes an iterative Q&A dialogue to select and collect phenotypes that optimize diagnostic value during the diagnostic session.
• Multi-Omics Differential Diagnosis: Integrates multi-omics data to refine candidate prioritization across molecular data types.

Scientific Applications:

• Enhanced Diagnostic Accuracy: In a diagnostic trial involving 238 published rare disease patients, RDmaster demonstrated superior performance compared to existing rare-disease diagnostic tools and large language models such as ChatGPT.
• Refined Multi-Omics Diagnosis: Applies multi-omics integration to improve differentiation among competing rare disease hypotheses.

Methodology:

Combines two Bayesian diagnostic methods, computes adaptive information gain using the AIGGI metric (including a Gini index component), and employs an iterative Q&A phenotype interrogation process with integration of multi-omics data.