Phenome risk classification enables phenotypic imputation and gene discovery in developmental stuttering

Area of Science:

• Genetics
• Speech and Language Pathology
• Biomedical Informatics

Background:

• Developmental stuttering is a speech disorder with a significant lifetime prevalence (6-12%).
• Electronic health records (EHRs) underdiagnose stuttering, with only 0.15% of individuals identified via diagnostic codes.
• Accurate identification of affected individuals is crucial for genetic research and understanding stuttering.

Purpose of the Study:

• To develop and validate a novel method for identifying individuals with developmental stuttering within the Vanderbilt EHR-linked biorepository (BioVU).
• To leverage this identification method for large-scale genetic association studies (GWAS) to uncover genetic variants linked to stuttering.
• To assess the clinical relevance of the findings through polygenic-risk prediction and concordance analysis.

Main Methods:

• Developed PheML, a PheCode-driven Gini impurity-based classification and regression tree model, to impute stuttering status using comorbidities.
• Applied PheML to BioVU data to identify a large cohort of individuals affected by stuttering.
• Conducted ancestry-stratified genome-wide association studies (GWAS) on PheML-imputed individuals and matched controls.

Main Results:

• PheML identified 9,239 genotyped individuals affected by stuttering in BioVU, revealing a clinical prevalence of approximately 10%.
• GWAS identified significant genetic variants associated with stuttering: rs12613255 near CYRIA (chromosome 2) in European ancestry and rs7837758 within ZMAT4 (chromosome 8) in African ancestry.
• Polygenic-risk prediction and concordance analysis validated the GWAS findings and demonstrated clinical relevance.

Conclusions:

• The PheML model effectively identifies individuals with developmental stuttering in EHR data, significantly increasing the number of cases for genetic research.
• Novel genetic variants associated with stuttering have been identified in European and African ancestries, contributing to our understanding of stuttering etiology.
• This population-based genetic analysis approach holds significant clinical relevance for stuttering research and potentially for future diagnostic or therapeutic strategies.