Potential and Considerations for Implementing Next-Generation Sequencing in Dutch Newborn Screening

Abstract

Newborn blood spot screening (NBS) is an important cornerstone of public health services, providing early detection of genetic disorders in newborns. With the dropping costs of next-generation sequencing (NGS), NGS is paving its way to be used in NBS. The implementation of NGS could be useful to complement the current biochemical tests because not all diseases can be traced through biochemical markers. Furthermore, the addition of new disorders is made less complex by not having to design and develop a biochemical assay. Unlike in diagnostics, where sequencing is already done more routinely, with screening, there is no prior information on the phenotype. Prior phenotype definition often allows for a more focused search in the genome, but in screening, no assumptions can be made about a possible disease. However, 100 NBS-related genes will be selected for a more targeted approach. In this study, we developed the VIPR (Variant Interpretation and Prioritization RIVM) pipeline to investigate the use of curated databases, population databases, and pathogenicity predictors to filter, interpret, and prioritize variants in NBS samples. In total, we collected 38 whole-exome sequencing (WES) samples: 23 Radboud patient samples, 14 Doetinchem cohort participants, and 1 RIVM volunteer as control samples. The workflow consists of three main processes. First, we filter variants based on depth (≥20), population allele frequency (≤0.05), and impact (moderate and high). Secondly, we classify variants based on curated databases ClinVar and VKGL, followed by prioritization with pathogenicity predictors. Lastly, the pathogenic variants will be further inspected on variant allele frequency and other mutations in the affected gene. This resulted in finding 21/23 pathogenic samples, achieving a high sensitivity of 91% (n=23, 95% CI ≈ 72%-99% Clopper-Pearson). VIPR correctly identified 8/15 of the healthy individuals, yielding a specificity of 47% (n=15, 95%, CI ≈ 21%-73% Clopper-Pearson). Several samples were misclassified as pathogenic (7/15), but additional criteria, such as checking heterozygous missense variants, could lower the number of false positives. However, this also risks increasing the number of false negatives because not all affected samples show clear causal variants and therefore will be missed if stricter filtering is applied. The combination of curated databases, population databases, and pathogenicity predictors ensures a well-supported mechanism for assessing SNVs. We also find that pathogenicity predictors are unreliable when used as a standalone and serve only as complementary support. Currently, the number of false positives does not encourage the use of NGS as a first tier, but used as a second or third tier could prove a powerful addition.