The AudioGene Translational Dashboard for Diagnosing Autosomal Dominant Nonsyndromic Hearing Loss: Phenotypic Data...

Area of Science:

• Genetics
• Bioinformatics
• Medical Informatics

Background:

• Autosomal dominant nonsyndromic hearing loss (ADNSHL) presents significant genetic heterogeneity, complicating diagnosis through traditional clinical methods.
• Existing computational tools for ADNSHL diagnosis often lack accuracy and interpretability, hindering the application of precision medicine.
• There is a critical need for advanced diagnostic tools that offer transparent and interpretable genotype-phenotype correlations for clinicians.

Purpose of the Study:

• To develop and evaluate the AudioGene Translational Dashboard, an interpretable clinical informatics tool.
• To enhance genotype-phenotype correlations for improved diagnostic decision-making in ADNSHL.
• To integrate machine learning models with interactive visualizations for transparent diagnostic support.

Main Methods:

• Developed the AudioGene Translational Dashboard, incorporating AudioGene v4 (multi-instance SVM) and v9.1 (ensemble methods) machine learning models.
• Integrated six interactive visualization tools, including audiometric profile plots and clustering analyses, to aid clinical interpretation.
• Designed the dashboard to address the "70/30" phenomenon, providing confidence indicators for diagnostic predictions.

Main Results:

• The AudioGene Translational Dashboard demonstrated a 74% likelihood of identifying the causative gene within the top 3 predictions, offering "green flag" or "red flag" diagnostic guidance.
• Interactive visualizations significantly improved clinicians' ability to interpret and correlate phenotypic data with predicted genetic outcomes.
• The tool enhanced diagnostic confidence and interpretability, despite acknowledging remaining uncertainty requiring interpretive context.

Conclusions:

• The AudioGene Translational Dashboard represents an advancement in clinical informatics for ADNSHL genetic diagnosis by combining explainable AI and interactive visualizations.
• The tool enhances clinical interpretability and diagnostic accuracy, supporting precision medicine in hearing loss.
• Future work will focus on improving class balance and incorporating user-customizable features for broader clinical applicability.