A genomic strategy for precision medicine in rare diseases: integrating customized algorithms into clinical practice

Abstract

Background: Despite the use of Next-Generation Sequencing (NGS) as the gold standard for the diagnosis of rare diseases, its clinical implementation has been challenging, limiting the cost-effectiveness of NGS and the understanding, control and safety essential for decision-making in clinical applications. Here, we describe a personalized NGS-based strategy integrating precision medicine into a public healthcare system and its implementation in the routine diagnosis process during a five-year pilot program.

Methods: Our approach involved customized probe designs, the generation of virtual panels and the development of a personalized medicine module (PMM) for variant prioritization. This strategy was applied to 6500 individuals including 6267 index patients and 233 NGS-based carrier screenings.

Results: Causative variants were identified in 2061 index patients (average 32.9%, ranging from 12 to 62% by condition). Also, 131 autosomal-recessive cases could be partially genetically diagnosed. These results led to over 5000 additional studies including carrier, prenatal and preimplantational tests or pharmacological and gene therapy treatments.

Conclusion: This strategy has shown promising improvements in the diagnostic rate, facilitating timely diagnosis and gradually expanding our services portfolio for rare diseases. The steps taken towards the integration of clinical and genomic data are opening new possibilities for conducting both retrospective and prospective healthcare studies. Overall, this study represents a major milestone in the ongoing efforts to improve our understanding and clinical management of rare diseases, a crucial area of medical research and care.

Keywords: Genetic diagnosis; Genomic medicine; Next generation sequencing; Precision medicine; Rare diseases; Research implementation.

Fig. 1

Overview of an NGS-based pilot program for genetic diagnosis of rare diseases in a public reference healthcare hospital. All recruited individuals came from one of the eight provinces of Andalusia and were included into the study according to their clinical manifestations or family history, through an electronic platform filled out by the clinician responsible for the genetic study request. After library preparation, sequencing and the automatic phase of data analysis, the alignment, quality control (QC) and variant files were uploaded to the personalized medicine module (PMM) for evaluation by the geneticist. Once the variant interpretation was finished, the geneticist could prepare a genetic report accessible to patients and their requesting clinician in the electronic health record. *Except for genes with variants previously reported as pathogenic, hypomorphic alleles and variants showing incomplete penetrance or variable expressivity. Abbreviations: ACMG class, American College of Medical Genetics and Genomics and Association for Molecular pathology classification; AD, autosomal dominant; AR, autosomal recessive; B, benign; CIP, conflicting interpretations of pathogenicity; CNVs, Copy Number Variations; Comp het, compound heterozygous; Cov, Coverage; GATK BP, Genome Analysis ToolKit Best Practices; Hem, hemizygous; Het, heterozygous; Hom, homozygous; Indv, individual ; LB, likely benign; LP, likely pathogenic; MAF, minor allele frequency; P, pathogenic; SNVs, single nucleotide variants; VAF, variant allele fraction; VUS, variant of uncertain significance; XL, X-linked; XLd, X-linked dominant; XLr, X-linked recessive. Created with BioRender.com

Fig. 2

Overview of the personalized medicine module (PPM) tool. (A) The PMM tool incorporates functionalities for genomic data management, SNV/Indels analysis and integration into the clinic. After primary and secondary data analysis, VCF, BAM and BAI files are uploaded into the PMM data analysis module, as well as sample QC and CNVs files. Once the study sample is selected, users can automatically apply both virtual panels and a set of filters, together with other complementary attributes to reach a manageable number of diagnostic variants. The custom combination of filters and prioritization settings can be saved for further analyses, as well as the variant-associated annotations and expert classifications to facilitate variant interpretation. Finally, the tool has the possibility to semi-automatically generate a variant report to be integrated into the patient’s electronic health record. (B) PMM tool screenshots showing the prioritization module interface and the detailed variant window. Created with https://BioRender.com

Fig. 3

Case distribution per disease category. Distribution of referred genetic tests by disease category based on clinical suspicion and considering 11 different major disease categories. Bars show the size of each case set grouped by disease category (A, orange bars) and subcategories (B, blue bars). For clarity, case sets with n < 5 have been omitted as well as disease categories represented by only one subcategory with n ≥ 5 (hearing and dermatological disorders)

Fig. 4

The use of virtual panels is a key step in variant filtering and prioritization. (A) Comparison of the mean number of candidate variants in randomly selected patients after applying the different automatic filtering steps with and without the use of virtual panels. (B) Boxplot diagram comparing the number of variants retained using our automatic filtering strategy with or without the application of different virtual panels, according to the number of genes included. T-test analysis was performed using the no-panel filtering strategy as the reference group. The application of any-sized virtual panels significantly reduced the number of variants to manually prioritize (p-value < 0,001). Abbreviations: CIP†, conflicting interpretations of pathogenicity with at least one pathogenic or likely pathogenic entry; MAF, minor allele frequency

Fig. 5

Graphical representation of the main genetic outcomes obtained for this strategy. (A) The genetic diagnosis rate for the whole approach and broken down for each of the versions of pRARE. (B) Treemap showing the major disease categories, in which each plot is scaled to represent the number of studied cases. Derm: Dermatological. (C) Percentage of cases with a full genetic diagnosis (positive), with variants of unknown significance consistent with their phenotypes (uncertain), with a monoallelic likely causative variant in an autosomal recessive gene (partial), and with no candidate variants (unsolved) for each major disease category both overall (“All”; saturated colors) and for the three pRARE versions (“D1”, “D2” and “D3”; blurred colors)

Fig. 6

Genetic heterogeneity for each of the main disease categories and prevalence of the mutated genes. “TOP50 Genes” refers to the set of genes that, together, harbor variants that could explain the phenotype for at least 50% of cases with detected variants of each disease category, with the exception of categories “Hearing” and “Endocrine”, in which only two genes were mutated in more than 1 patient resulting in a prevalence < 50% (47% and 33% of cases, respectively). (A) Most prevalent mutated genes (TOP50 Genes) in our cohort of patients, showing the recurrence of each of them. The figure also illustrates the prevalence of these genes in different categories, when applicable. (B) Depiction of the percentage of cases with variants both in TOP50 Genes and the remained genes for each category (bar chart). The line graphs show the number of genes making up the TOP50 and the number of remained mutated genes per category, illustrating the genetic heterogeneity for each category, which is directly proportional to the distance between the points of both gene groups

Fig. 7

Cases included in observational studies and benefiting from treatments/therapies. Patients with positive diagnostic findings and/or their families were offered to be included in additional studies including carrier screening, presymptomatic, prenatal or preimplantational genetic testing. As a result, a total of 3121 individuals received reproductive genetic counselling, surveillance follow- up or genetic-guided therapeutic decisions. Also, a set of patients were recruited for observational studies based on the molecular results and 17 patients had access to a personalized pharmacological or gene therapy treatment. POS: Positives; NEG: Negatives