Clinician-centric diagnosis of rare genetic diseases: performance of a gene pertinence metric in decision support for clinicians

Michael M Segal,Ayman W El-Hattab,Kiely N James,Renee George,Peter Waltman,Joseph Gleeson,Valentina Stanley

doi:10.1186/s13023-020-01461-1

Abstract

BackgroundIn diagnosis of rare genetic diseases we face a decision as to the degree to which the sequencing lab offers one or more diagnoses based on clinical input provided by the clinician, or the clinician reaches a diagnosis based on the complete set of variants provided by the lab. We tested a software approach to assist the clinician in making the diagnosis based on clinical findings and an annotated genomic variant table, using cases already solved using less automated processes.ResultsFor the 81 cases studied (involving 216 individuals), 70 had genetic abnormalities with phenotypes previously described in the literature, and 11 were not described in the literature at the time of analysis (“discovery genes”). These included cases beyond a trio, including ones with different variants in the same gene. In 100% of cases the abnormality was recognized. Of the 70, the abnormality was ranked #1 in 94% of cases, with an average rank 1.1 for all cases. Large CNVs could be analyzed in an integrated analysis, performed in 24 of the cases. The process is rapid enough to allow for periodic reanalysis of unsolved cases.ConclusionsA clinician-friendly environment for clinical correlation can be provided to clinicians who are best positioned to have the clinical information needed for this interpretation.

Highlights

In diagnosis of rare genetic diseases we face a decision as to the degree to which the sequencing lab offers one or more diagnoses based on clinical input provided by the clinician, or the clinician reaches a diagnosis based on the complete set of variants provided by the lab
Because the clinical information is typically submitted before the genomic sequencing, there is little opportunity for the lab to bring into the clinical correlation information prompted by the unusual gene variants found in sequencing. Another approach is for clinicians to remain at the center of clinical correlation, drawing on the detailed clinical characterizations in the primary literature as well as comprehensive reviews in the literature. Such clinical correlation is difficult for clinicians to perform because of the complexity involved in dealing with thousands of variants and thousands of diseases
We examine here whether we could use copy number variants (CNVs) and single nucleotide variants (SNVs), both derived from genomic sequencing, in a single test, which could offer a process for clinical correlation that is more efficient

Summary

Introduction

In diagnosis of rare genetic diseases we face a decision as to the degree to which the sequencing lab offers one or more diagnoses based on clinical input provided by the clinician, or the clinician reaches a diagnosis based on the complete set of variants provided by the lab. One approach is to submit DNA to labs to use massively parallel sequencing to identify genetic diagnoses, but this approach is difficult to implement in an optimal way because clinicians provide limited information to Another approach is for clinicians to remain at the center of clinical correlation, drawing on the detailed clinical characterizations in the primary literature as well as comprehensive reviews in the literature. Such clinical correlation is difficult for clinicians to perform because of the complexity involved in dealing with thousands of variants and thousands of diseases. An early version of the system was tested in the CLARITY genome-analysis competition and was the most clinician-centric of the analyses and it performed by far the fastest [3]

Methods

Results

Discussion

Conclusion