2023 ACMG Annual Clinical Genetics Meeting Digital Edition: Poster Gallery-Anonymous is Synonymous: Utilizing De-identified Genetic and Clinical Information to Evaluate Diagnostic Prioritization Algorithms

Pdf Summary

The study explores the use of de-identified patient data to validate diagnostic prioritization algorithms for genetic diagnosis. The cohort consisted of 93 patients diagnosed by exome or genome sequencing. De-identified variant call format (VCF) files were created for each patient by inserting their causative variant(s) into publicly available VCFs. A note containing phenotype information for each patient was selected and analyzed to extract phenotypes. The original and de-identified sequencing data was referred to as IDSeq and AnonSeq, respectively, and the original and de-identified notes were referred to as IDNote and DeIDNote, respectively.<br /><br />The diagnostic variant prioritization algorithm CAVaLRi was used with all combinations of identified and de-identified information sources. The average diagnostic variant rank (ADR) was calculated for each combination. The results showed that de-identifying notes did not affect ranking for original sequencing data, but using de-identified VCFs did affect the rank. The difference in ADR between IDSeq-IDNote and AnonSeq-DeIDNote was 3.14 ranks.<br /><br />The study concludes that de-identifying the genetic representation significantly affected the ADR, but the note de-identification procedure did not. If these effects can be replicated with other diagnostic prioritization algorithms, this data set and approach could accelerate algorithmic development. Future work should focus on streamlining the note de-identification process and improving methods to create anonymous VCFs. The de-identified data set is available for sharing upon request. The study includes figures showing the comparison of average diagnostic rank for combinations of identified and de-identified data, the percentage of cases solved at a given diagnostic variant rank, and PR curves comparing CAValRi outputs at a given threshold.

Asset Subtitle

Presenting Author - Brandon Stone, MD; Co-Author - Robert J. Schuetz, BS; Co-Author - Austin A. Antoniou, PhD; Co-Author - Emma Garval, BS; Co-Author - Amad Hussain, BS; Co-Author - Bimal P. Chaudhari, MD MPH;

Meta Tag

Bioinformatics

Exome sequencing

Genetic Testing

Genome sequencing

Genomic Methodologies

NextGen Sequencing

Phenotype

Sequencing

Variant Detection

Co-Author Robert J. Schuetz, BS

Co-Author Austin A. Antoniou, PhD

Co-Author Emma Garval, BS

Co-Author Amad Hussain, BS

Co-Author Bimal P. Chaudhari, MD MPH

Presenting Author Brandon Stone, MD

Keywords

de-identified patient data

diagnostic prioritization algorithms

genetic diagnosis

exome sequencing

genome sequencing

variant call format (VCF)

phenotype information

CAVaLRi algorithm

diagnostic variant rank

note de-identification