2023 ACMG Annual Clinical Genetics Meeting Digital Edition: Poster Gallery-Assessing the Performance of Automated HPO Term Extraction for Deep Phenotyping of Patients Receiving WES/WGS in a Clinical Diagnostic Laboratory

Pdf Summary

The document discusses the use of Natural Language Processing (NLP) algorithms to automate the extraction of Human Phenotype Ontology (HPO) terms from patient clinical narrative notes. This process is important for creating a prioritized gene list for whole exome and whole genome sequencing (WES/WGS) analysis. The document presents the results of a study that assessed the performance of an NLP-assisted deep phenotyping and gene-ranking workflow in a clinical diagnostic laboratory setting.<br /><br />The study included 50 pediatric patients who underwent WES, and their clinical notes were processed using the EHR-Phenolyzer pipeline, which utilized NLP to extract HPO terms and generate a ranked gene list. The accuracy of the extracted HPO terms was compared to provider-submitted terms and other workflows.<br /><br />The results showed that the NLP-assisted workflow efficiently translated phenotypes into HPO terms and outperformed other workflows in ranking disease-causing variants at certain positions. On average, 27.8 HPO terms were extracted per case compared to an average of 7.3 features manually submitted by providers.<br /><br />Applying the optimized pipeline to 20 WGS datasets, the ranking system was able to prioritize genes with candidate disease-causing non-coding/intronic variants in 3 cases at positions <150th. The study also discovered 3 cases with pathogenic coding variants in genes that were poorly ranked, but these cases had atypical presentations or emerging evidence for the genes in the literature.<br /><br />Overall, the study demonstrates that NLP-assisted deep phenotyping can efficiently convert phenotypic features into HPO terms and improve the accuracy of gene-ranking algorithms. This method can be valuable in prioritizing non-coding/intronic variants in patients with previously undiagnostic WES results. Further human curation can further improve the accuracy of the extracted HPO terms.

Asset Subtitle

Presenting Author - Kai Lee Yap, PhD; Co-Author - Anthony Wong, PhD; Co-Author - Sachleen Tuteja, NA; Co-Author - Andrew Skol, PhD; Co-Author - Andy Drackley, MS, CGC; Co-Author - Alexander Ing, MS, CGC; Co-Author - Eleanor Hilton, MS, CGC; Co-Author - Michael Carroll, PhD; Co-Author - Christopher McCabe, BSc; Co-Author - Sabah Kadri, PhD; Co-Author - Pamela A. Rathbun; Co-Author - Katrin Leuer, PhD; Co-Author - Joel Charrow, MD;

Meta Tag

Exome sequencing

Genetic Testing

Genome sequencing

Genomic Methodologies

Genotype-Phenotype Correlations

Identification of Disease Genes

Methodology

NextGen Sequencing

Phenotype

Phenotypic delineation of disorders

Sequencing

Co-Author Anthony Wong, PhD

Co-Author Sachleen Tuteja, NA

Co-Author Andrew Skol, PhD

Co-Author Andy Drackley, MS, CGC

Co-Author Alexander Ing, MS, CGC

Co-Author Eleanor Hilton, MS, CGC

Co-Author Michael Carroll, PhD

Co-Author Christopher McCabe, BSc

Co-Author Sabah Kadri, PhD

Co-Author Pamela A. Rathbun

Co-Author Katrin Leuer, PhD

Co-Author Joel Charrow, MD

Presenting Author Kai Lee Yap, PhD

Keywords

Natural Language Processing

NLP

Human Phenotype Ontology

HPO terms

patient clinical narrative notes

gene list

whole exome sequencing

whole genome sequencing

NLP-assisted deep phenotyping

gene-ranking workflow