Sample preparation.

Nurses perform the heel-prick test on all newborns between 24 and 72 hours after birth as part of the tNBS program. This involves collecting a small blood sample from the infant’s heel and depositing it onto DBS cards to fill the standard four circles, ensuring that the blood soaks through the paper completely. Once room temperature-dried, the DBS cards are transported to the Laboratory of Metabolic Diseases Unit (UDyTEMC) at the Clinical Hospital of Santiago de Compostela, following the tNBS workflow. No additional blood will be drawn for the study, as residual blood from the routine screening will be used, which will be sufficient for WES. The case manager will access the remaining sample using the internal code and will punch out three 3.2-mm discs using a biopsy punch. The discs will be placed in a 1.5 mL Eppendorf Safe-Lock DNA LoBind Tube (Eppendorf) and stored at 4°C until DNA extraction.

DNA will be extracted from DBS using the Chemagen kit (PerkinElmer). Briefly, samples will undergo lysis and protein digestion, followed by magnetic bead–based purification and elution in nuclease-free water. DNA concentration and purity will be quantified using both spectrophotometric (NanoDrop) and fluorometric (PicoGreen) assays. Extracted DNA will be stored at 4 °C when processing and shipment are scheduled on different days and will subsequently be delivered to the FPGMX genomic sequencing platform for WES.

An aliquot of the extracted DNA will be transferred to the genomic sequencing platform at FPGMX, where it will receive a unique internal identification code. Library preparation will be performed using the KAPA HyperPrep Kit, followed by target enrichment with KAPA HyperExome probes according to the manufacturer’s recommendations. Libraries will be sequenced on an Illumina NovaSeq 6000 system using 2 × 100 bp paired-end reads, achieving a mean coverage of 100–200× across coding regions and flanking intronic sequences. Bioinformatic processing will be conducted by FPGMX. Reads will be aligned to the human reference genome (hg38), and single-nucleotide variants (SNVs), insertions/deletions (indels), and copy number variants (CNVs) will be identified and annotated using established pipelines incorporating DRAGEN, Picard, mosdepth, bedtools, and Haplogrep. Variant interpretation will be restricted to genes included in the virtual NBGenS panel.

CNV candidates will be generated using the provider’s DRAGEN germline CNV pipeline (v4.2.7) [42], with copy-number calling enabled and panel-of-normals normalization, together with an in-house GATK-gCNV workflow based on a cohort model [43]. For GATK-gCNV, read-count intervals will correspond to the exome-capture targets and will be generated using PreprocessIntervals with a bin length of 0 and 250-bp padding. GATK-gCNV calls will be retained at a quality score ≥100 and DRAGEN calls at FILTER PASS. Two calls will be considered concordant, and therefore a high-confidence candidate, when they show ≥50% reciprocal overlap and affect the same gene in the same dosage direction (loss or gain); calls supported by a single caller will be reviewed but not classified as high-confidence. Calls larger than 3 Mb will be discarded as capture-driven artefacts, leaving an effective resolution down to a single captured target or exon.

The analytical performance of the combined workflow will be evaluated before clinical reporting to establish the validated minimum reportable range, particularly for single-exon events. CNV calling will be performed across the captured exome regions, but clinical interpretation and reporting will be restricted to genes included in the virtual panel. Off-panel CNVs will not be returned to families. Analytical performance will be estimated against reference materials with known copy-number states and previously array- or MLPA-confirmed in-house cases, reporting sensitivity and positive predictive value stratified by event size to define the minimum reportable range.

Panel CNVs will be classified according to the 2019 ACMG/ClinGen constitutional CNV framework [44], supported by ClassifyCNV [45], AnnotSV, and ClinGen dosage-sensitivity curations. Intragenic single- and multi-exon CNVs will be assessed according to gene dosage sensitivity and their predicted effect on the reading frame.

Concordant calls from both approaches will be considered high-confidence candidates. Clinically actionable CNVs will be confirmed using Multiplex Ligation-dependent Probe Amplification (MLPA) or chromosomal microarray analysis before communication to the family. The limitations of WES-based CNV detection, particularly for small, single-exon, intronic, or poorly captured events, will be considered when interpreting negative results.

A containment protocol will be applied in cases of technical issues. When sufficient material is available, the affected step will be repeated; if necessary, earlier stages of the protocol will also be revisited. Close coordination between the sequencing and research teams will be maintained to support turnaround times consistent with the projected schedule.

Gene panel.

The gene panel was selected according to the following criteria: i. gene–disease association (strong), ii. age of disease onset (<15 years), iii. effectiveness of early treatment (significance), and iv. severity and penetrance of the condition (severe and high).

Starting from an initial list compiled from the literature, existing panels, and other relevant sources, gene–disease pairs were classified into three groups: (i) pairs that adequately met all predefined inclusion criteria; (ii) pairs that met these criteria but required consideration of specific interpretative or clinical conditions, such as incomplete penetrance or variable expressivity; and (iii) pairs associated with conditions also detectable through conventional biochemical newborn screening. The detailed allocation of gene–disease pairs to these three groups is not shown in S3 Table, as this classification will remain subject to ongoing curation during subsequent panel updates. This classification will support documentation and periodic review of the panel but will not, by itself, determine whether a finding is reportable. A core panel was established from these groups and will be periodically updated as new evidence emerges, allowing gene–disease pairs to be added, removed, or reclassified. This dynamic-panel approach will ensure flexibility and scalability while preserving the scientific consistency and predefined inclusion standards of the panel (S3 Table). The current version comprises 549 genes and 603 gene–disease pairs.

Variant analysis and interpretation.

Variant interpretation and classification will follow the recommendations of the ACMG [32], the ACGS Best Practice Guidelines for Variant Classification in Rare Disease 2024 [46], and the Sequence Variant Interpretation (SVI) general recommendations for applying ACMG/AMP criteria [36].

Once sequencing has been successfully completed, the research team will be notified by the FPGMX to initiate the analytical process for the resulting variants using an internally developed FPGMX tool based on R Shiny and HPC scripting. This tool will allow all sequencing variants from a project to be consolidated, will keep their clinical and functional annotation up to date, and will enable rapid querying through a web interface that will communicate with the IGV genome viewer (referred to as the VarDB analysis platform).

To optimize variant interpretation, a set of restrictive criteria will be applied within the VarDB analysis platform to streamline the process (Table 2). Potential variants will be extracted for each participant and analyzed considering to their zygosity and the inheritance pattern of the disease selected in the panel. Although VUS may be observed, only pathogenic or likely pathogenic variants will be considered for subsequent analyses. VUS will be reserved solely for exploratory UM, which will be used as a functional assay to assess the pathogenicity of the variant.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Restrictive criteria applied in VarDB for variant characterization.

https://doi.org/10.1371/journal.pone.0359270.t002

For variants associated with recessive diseases, the presence of additional variants in the same gene will be assessed to identify possible compound heterozygosity. In both scenarios -recessive variants (either in compound heterozygosity or homozygosity) and dominant variants (in heterozygosity or homozygosity)- only potentially reportable findings (PRF) will be considered. These findings will need to be supported by clinical evidence from major information sources, including Franklin [47], VarSome [48], OMIM [40], ClinVar [49], ClinGen [50], and PubMed [51].

Variant interpretation will be conducted in pairs by the research team, and results will be compared to reach a consensus conclusion. If variants of interest (referred to as PRF), are identified, the suspicious case protocol will be activated, and meetings will be convened with clinicians who are experts in the relevant disease category.

The VarDB analysis platform will use InterVar (2015) [52] to assign ACMG classifications to variants. Given this potential limitation, variant classifications will also be cross-checked against more up-to-date databases, such as Franklin.

A scheduled case reanalysis will be performed 12 months after the initial analysis. Thereafter, the virtual panel, gene–disease associations, interpretation resources, and relevant clinical recommendations will be reviewed annually throughout the data-retention period. Reanalysis may also be initiated at any time following new clinical, biochemical, metabolomic, or phenotypic information; a material update to the panel or bioinformatic workflow that may affect diagnostic sensitivity; new robust gene–disease or variant–disease evidence; new clinical recommendations; or reclassification of a previously identified variant with potential implications for diagnosis, prevention, surveillance, or treatment.

Families will be recontacted only when reanalysis identifies a new or modified result that meets the protocol reporting criteria and has potential clinical relevance. Reanalyses without clinically relevant changes will not be communicated individually, although their completion and outcome will be documented. For further details on the reanalysis triggers and family recontact criteria, please refer to S4 Fig.

Standard WES will not be considered reliable for the definitive detection of pathogenic repeat expansions. As a targeted complementary analysis, panel genes containing known pathogenic repeat regions will be screened using ExpansionHunter v5 and a curated variant catalogue applied to the WES reads. These calls will be treated as screening-level findings only. If clinical findings or family history suggest a repeat expansion disorder, such as Friedreich ataxia, fragile X syndrome, or myotonic dystrophy, an appropriate targeted assay, including repeat-primed PCR or Southern blotting, will be considered irrespective of the WES findings.

Mitochondrial DNA variants will only be assessed at positions meeting predefined depth and quality requirements because mitochondrial genome coverage from exon-capture WES will be variable. If mitochondrial disease is clinically suspected, dedicated mitochondrial DNA sequencing will be considered rather than relying on WES alone. Mitochondrial variants will be reported only at positions with a minimum read depth of 100× and base and mapping quality ≥20, and heteroplasmy will be quantified and reported at variant-allele fractions ≥10%.

Variants located near captured splice sites will be evaluated using SpliceAI and Pangolin with predefined thresholds. However, most deep intronic pathogenic variants will not be detected because they fall outside the captured regions. Genome sequencing or RNA-based testing may therefore be considered when clinical suspicion remains high despite a negative WES result. The predefined thresholds will be a SpliceAI or Pangolin delta score ≥0.20 as supporting evidence of a splicing effect, ≤ 0.10 as evidence against, and 0.10–0.20 as uninformative; for canonical splice-site positions without an exact delta score, a dbscSNV ada or rf score >0.6 will be used as a fallback.

Mosaic-sensitive tools, including DeepMosaic and MosaicForecast, will be used to support the identification of low variant-allele-fraction candidates. Nevertheless, low-level mosaic variants may remain below the detection threshold, and analysis of blood will not exclude tissue-restricted mosaicism. This analysis will target variant-allele fractions of approximately 1–30%, will be restricted to genes with neonatal-actionable mosaic phenotypes, and will require manual inspection of every candidate.

Repeat expansions, mitochondrial variants, deep intronic variants, and low-level mosaic variants will consequently remain underdetected by WES. Diagnostic yield estimates for these variant classes will therefore be interpreted as lower-bound estimates, and cases with strong clinical suspicion and negative or inconclusive WES findings may undergo appropriate targeted or orthogonal testing.

Genetic variants report.

Each case is individually reviewed by a multidisciplinary committee and discussed with the responsible clinician before disclosure, after which a report is prepared and issued by clinical molecular geneticists from the Laboratory of Metabolic Diseases Unit. The structure and content of reports are adapted from genomic reports routinely developed following the ACMG recommendations [32]. These reports consist of a single document: the first page includes the delivery address of the family residence, participant details, a “Results Summary” outlining the findings, followed by an “Interpretation Summary,” and a final “Summary of the Type of Analysis Performed.” The second page includes Annex I (list of genes analyzed in the NBGenS virtual panel) and Annex II (gene coverage specifications).

Following the completion of the analytical and interpretative process for each participant’s variants, cases are classified into one of two possible groups: No reportable findings (NRF) or Potentially reportable findings or PRF. Classification as PRF requires the identification of variants that meet the criteria for reporting-namely, pathogenic, or likely pathogenic variants in genes supported by scientific evidence demonstrating pediatric-onset disease causality for which early intervention is possible in a zygosity compatible with the inheritance pattern of the disease. For conditions showing variable expressivity, reporting decisions are based primarily on the evidence available for the specific variant and, when this is insufficient, on the corresponding genotype and gene–disease association. All other scenarios are classified as NRF, including the absence of reportable pathogenic or likely pathogenic variants; VUS, even when selected for untargeted metabolomics or subsequent reevaluation; isolated heterozygous pathogenic or likely pathogenic variants in genes associated with autosomal recessive diseases; and pathogenic or likely pathogenic variants associated with low, incomplete, variable, or insufficiently established penetrance, unless they are considered clinically relevant and enable preventive, surveillance, follow-up, or therapeutic measures with potential benefit for the child. These findings will not be communicated as clinically positive results, although they may be retained for subsequent research evaluation.

ACMG secondary findings are neither actively sought nor reported. Variant filtering, interpretation, and reporting are restricted to the predefined NBGenS disease and pharmacogenetic panels; variants outside these analytical scopes are excluded from review and therefore cannot be returned as secondary findings within the NBGenS workflow. Any subsequent analysis beyond the original panel prompted by the development of a relevant phenotype will be considered a separate clinically indicated investigation and managed according to the applicable clinical, ethical, consent, and ACMG secondary-findings procedures. For further details on the variant interpretation and reporting decision framework, please refer to S4 Fig.

For participants classified as NRF, reports are sent by regular postal mail following standard tNBS procedures. For suspicious cases with a PRF, a specific protocol is activated, and the report includes an additional summary of the variant of interest. These reports are never mailed directly to the family. Instead, each case is reviewed in interdisciplinary meetings with the relevant clinical specialists, after which parents are informed of the findings and invited to an in-person consultation with pediatricians and clinical geneticists to receive the results, genetic counseling, and appropriate follow-up. All parent-directed communications (phone calls and in-person visits) follow the best practices described by Remec et al. [20].

To streamline report generation, a Visual Basic for Applications (VBA)-based macro is employed for automation. The initial Visual Basic for Applications code is developed with the aid of artificial intelligence but undergoes review and refinement by expert bioinformaticians (S5 File).

Pharmacogenetic study.

The pharmacogenetic study will constitute a secondary objective of the NBGenS project. It will aim to determine the detection rate of variants in genes with strong pharmacological associations relevant to current or potential paediatric treatments and to assess their possible inclusion in a gNBS programme. Compared with the primary objective, this analysis will differ in the selection of genes, the variant analysis and interpretation strategy, and the reporting of results.

The initial gene–drug list will be compiled from PharmVar [53], the Centers for Disease Control and Prevention (CDC) Tier 1 genomic applications [54], the AEMPS Pharmacogenomic Biomarker Database [55], and PharmGKB databases [56]. The resulting research universe will comprise 1,160 gene–drug associations involving 81 genes and 464 drugs, as detailed in S6 Table.

Star alleles will be inferred from aligned WES CRAM files using PyPGx v0.26 for the pharmacogenes supported by the software. Statistical haplotype phasing will be combined with copy-number estimation based on normalised read depth for CNV-defined loci, including CYP2D6 and UGT1A1. Genes not resolved using PyPGx will be diplotyped from the joint-genotyped VCF using PharmCAT v3.2.0, while HLA-B*57:01 and HLA-A*31:01 will be typed using HLA-LA.

Pharmacogenetic findings will be organised into clinical and exploratory tiers. The clinical tier will be restricted to CPIC Level A actionable gene–drug pairs involving 22 genes for which prescribing guidance from CPIC, DPWG, or the FDA is available. Within this tier, priority will be given to associations with PharmGKB evidence levels 1A–2A and to AEMPS-authorised medicines with paediatric relevance. Associations with lower levels of evidence, without paediatric relevance, or involving medicines not covered by the AEMPS criteria will be retained for possible future reanalysis but will not be returned under this protocol.

The recognised limitations of exon-capture WES at complex pharmacogenetic loci will be considered during interpretation. In particular, WES may not reliably resolve CYP2D6–CYP2D7 hybrid alleles, whole-gene deletions or duplications, or pseudogene interference, and rare CYP2C19 structural variants may be missed. If a reliable CYP2D6 copy-number or structural call cannot be obtained, no clinically actionable structural result will be issued from WES data alone.

All clinically actionable pharmacogenetic findings will undergo orthogonal confirmation using an appropriate targeted method, such as a TaqMan copy-number or genotyping assay, long-range PCR, Sanger sequencing, or a validated pharmacogenetic genotyping panel, before communication. A healthcare professional with training in pharmacogenetics will communicate these findings to families through a separate report, which will be incorporated into the electronic health record of the Galician Health Service. Findings with insufficient clinical relevance, analytical reliability, or actionability will not be returned.

This secondary analysis will highlight the considerable versatility and potential utility of the genomic information that will be collected at the individual level, enabling both alternative analytical approaches and future reanalysis that may prove clinically valuable should new indications arise during the study period.

Untargeted metabolomics study.

After completing tNBS tests and WES, variants will be analyzed according to the previously described criteria (Table 2) using the VarDB analysis platform. Because available resources and analytical capacity do not permit untargeted metabolomic evaluation of every VUS, variants will be prioritized for this analysis. VUS prioritized for untargeted metabolomics are defined as variants that, following initial classification according to ACMG/AMP and ClinGen recommendations, could potentially reach the likely pathogenic category if metabolomic findings provided additional applicable phenotypic or functional evidence, such as PP4 or PS3. The analytical criteria that will be used to determine participant eligibility for UM analysis are summarized in Table 3.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Inclusion and exclusion criteria for NBGenS participants in untargeted metabolomics.

https://doi.org/10.1371/journal.pone.0359270.t003

Participants meeting any of the inclusion criteria will have their samples prepared for shipment to the designated reference laboratory (Hospital Clinic de Barcelona, Spain). To preserve pseudonymization, all samples will be identified exclusively by their NBGenS study identification code.

Correctly labelled samples will be organized using a standardized system of envelopes and protective bags before shipment. Metabolites will be extracted from the DBS samples using a water/methanol solution. After evaporation of the eluate, the residue will be reconstituted with water containing 0.1% formic acid and injected into a Shimadzu LC40 ultra-performance liquid chromatography (UPLC) system coupled to a high-resolution time-of-flight mass spectrometer (X500R-QTOF, Sciex S.A). Chromatographic separation will be performed using an Acquity HSS T3 column (2.1 × 100 mm, 1.8 μm; Waters, Milford, MA, USA).

Data analysis will be performed using SCIEX OS software in combination with an in-house–developed automated bioinformatic pipeline. Metabolite levels in each participant will be compared with those observed in a reference control cohort using Z-scores, and values ≥ +2 or ≤ −2 will be considered potentially abnormal. The pipeline will perform data processing, Z-score calculation, and the initial identification of altered metabolites. All metabolomic profiles will subsequently be reviewed by a multidisciplinary team with expertise in clinical metabolomics and inherited metabolic disorders. This review will consider potential confounding factors, including prematurity, gestational age, nutritional status, concomitant medical treatments, intercurrent illness, and metabolic patterns previously described in these settings. Only abnormalities consistent with the pathophysiological mechanism of the disorder under evaluation and not reasonably explained by these potential confounders will be considered supportive evidence for clinical interpretation. The reference control cohort will comprise unaffected newborns analysed on the same platform and matched for gestational and postnatal age, feeding, and analytical batch. Because a |Z| ≥ 2 threshold will flag a proportion of features by chance across the untargeted profile, an isolated Z-score will not be treated as evidence in itself: only alterations that are biologically consistent with the candidate variant and reproducible will be retained, and no metabolite will contribute as ACMG evidence on the basis of the Z-score alone.

The data obtained will be systematically recorded in a Microsoft Excel database. Prior to metabolomic analysis, the dataset will include each participant´s unique study identification code and variants of interest, including the gene, OMIM entry, inheritance pattern, zygosity, NM accession, variant type, ACMG classification, specific comments, and timestamps for the different procedural stages.

Once UM results are available, the following information will be documented: metabolomics outcome (normal, abnormal, or insufficient sample), metabolite name and type of alteration (increase or decrease), potential correlation between the biochemical alteration and the identified variant, current ACMG classification, and any proposed reclassification.

This complementary approach will enable the identification of biochemical abnormalities associated with variants of uncertain or incompletely characterized clinical significance and will provide additional evidence for their interpretation. In cases where a VUS is identified together with an untargeted metabolomic profile compatible with the disorder under evaluation, the metabolomic finding will be considered an additional source of evidence within the structured ACMG/AMP classification framework but will not, in isolation, determine variant reclassification or clinical reporting. Its applicability as phenotypic or functional evidence—for example, PP4 or PS3, respectively—and the strength assigned will be determined according to the disease mechanism, the specificity of the metabolic profile, and the degree of analytical and clinical validation of the assay, in accordance with ACMG/AMP and ClinGen recommendations. Each case will be reviewed by a multidisciplinary committee that will integrate the metabolomic findings with population, computational, segregation, phenotypic, functional, and case-level evidence. A variant will be reclassified as likely pathogenic and clinically communicated only when the combined independent evidence meets one of the ACMG/AMP combinations established for that category, no relevant contradictory evidence is present, and the finding meets the reporting criteria defined in the protocol. Variants that do not meet this threshold will remain classified as VUS, will not be communicated as clinically positive findings, and may undergo follow-up or further analysis exclusively within the research context.

In addition, UM will allow the exploration of a much wider range of biochemical alterations than tNBS, enabling the detection of metabolic disturbances linked to a broader spectrum of diseases. It will also offer a broad range of measurable biochemical features that may be leveraged for biomarker discovery.

Sequencing the entire exome will provide an additional advantage: when a metabolic alteration is detected that does not correlate with previously identified variants, the exome may be retrospectively interrogated to perform a targeted search across all known genes potentially implicated in the observed biochemical abnormality. These integrated findings may support the formulation of hypotheses beyond tNBS and provide deeper insight into individualized risk assessment and metabolic characterization. However, the integration of these approaches will also represent a significant methodological and implementation challenge; therefore, their added clinical value, feasibility within the screening framework, and optimal strategies for communicating results will need to be carefully evaluated.

Cost-effectiveness study.

A Markov decision-analytic model will be developed to perform both cost-effectiveness and cost-utility analyses, estimating the discounted incremental cost per life-year gained and per quality-adjusted life-year (QALY) gained for neonatal genetic screening compared with no screening. The analysis of the economic implications of gNBS with particular emphasis on cases in which a PRF is identified and classified as suspicious following clinical consensus, will incorporate key parameters including disease prevalence, variant allele frequency, degree of associated disability, diagnostic test reliability, comparisons between early and delayed intervention, the preventive potential of early detection, available therapeutic options, and the potential benefits for relatives in whom familial segregation may be identified.

Estimates of the expected net changes in costs and life expectancy derived from genetic screening will be used to calculate incremental cost-effectiveness ratios. Sensitivity analyses of the main variables will be performed using Monte Carlo simulations. The economic evaluation will apply discount rates to both costs and health outcomes, with a base case of 3% and a range of 0% to 5% for sensitivity analyses. In addition to the cost-effectiveness analysis, a budget impact analysis will be conducted, given the need to estimate the financial consequences of adopting or expanding a new healthcare technology or intervention within a specific context [57].

AI-assisted language editing.

During the preparation of this manuscript, the authors used ChatGPT (OpenAI) to assist with language editing and improvement of the text and to generate the non-data-bearing illustrative images used in the design of S2 Fig, based on prompts written by the authors. The tool was not used for data analysis, interpretation, or generation of scientific content. All generated content was carefully reviewed and edited by the authors, who take full responsibility for the final version of the manuscript.