Approved Research

Distance-based Panel Generation Optimizes Gene Selection for Targeted Gene Panel Design

Lay summary

Genetic variants are the cause of many genetic diseases. When studying a specific disease, gene panels, a collection of genes sharing an association with that disease, are often used to help the researcher focus only on genetic variants affecting those genes. The process of gene panel generation presents a challenge: on the one hand, adding less genes to the panel means reducing the number of identified candidate genetic variants requiring interpretation. On the other hand, limiting the number of genes included in the panel may result in exclusion of important and relevant genes, therefore missing the causative genetic variant. Currently, gene panels for various diseases are offered by numerous clinical laboratories. These panels differ significantly from one another and selecting the most appropriate gene panel for a given disease is often challenging. There is a need for a gene panel design method that identifies the most relevant genes while limiting the number of genes that have no association with the clinical condition. We developed a method that generates a panel based on a combination of all publicly available panels. The association of each gene in the panel and the disease is scored based on the number of panels the gene appears in and additional metrics. Utilizing this score a gene panel can be designed to be extremely specific to the disease or very sensitive (including candidate genes with only a mild association). We compare the performance of our method against an established expert panel source called PanelApp and show that our method improves sensitivity while maintaining specificity comparable to the expert panel. Our approach offers researchers and clinicians a useful tool for designing gene panels while considering specific needs of sensitivity and specificity. Additionally, our method facilitates gene panel design while leveraging the collected wisdom of the entire set of publicly available panels. As the use of gene panels become more and more common, we predict our method will have a high impact in the field.