Comprehensive and rapid detection of genomic alterations in pediatric leukemias using whole-genome sequencing with adaptive sampling

Abstract Molecular and cytogenetic classification is essential for risk stratification and therapeutic decisions in pediatric leukemias. Conventional workflows typically require multiple tests, some offering rapid but focused insights (e.g., FISH), others providing broad information at the cost of longer turnaround times (e.g., karyotyping, exome, RNA-seq). This iterative and fragmented approach increases costs, delays decisions, and consumes limited tumor material. Whole-genome sequencing (WGS) with Oxford Nanopore Technologies (ONT) captures both genomic and epigenomic information, interpretable in real time. However, the coverage of a single flow cell (40X) limits sensitivity for subclonal events. ONT supports Adaptive Sampling (AS-WGS), which selectively enriches regions of interest to increase coverage, but workflows are still being optimized. We hypothesized that optimized AS-WGS could provide high coverage sufficient to detect subclonal alterations while retaining pan-genomic breadth, enabling the identification of all relevant genomic events in a single test to support rapid classification of pediatric leukemias. We applied an optimized AS-WGS protocol to 31 samples from 30 pediatric patients enrolled in the Signature research program. The cohort included 20 leukemia/MDS cases (8 B-ALL, 7 AML, 3 T-ALL, 1 Burkitt leukemia, 1 MDS), and 11 controls (7 solid tumors and 4 normal tissues). Genomic DNA was sheared (~12 kb) and sequenced using ONT SQK-LSK114 on FLO-PRO1114M flow cells, with adaptive sampling of a custom list of 380 genes/loci. Sequencing ran for 72 hours. Performance was benchmarked against alterations identified by extensive clinical testing (FISH, karyotyping, microarrays, exome and RNA-seq). Mutations with VAF 5% from clinical exomes were considered; alterations not called by the workflow were included if supported by ≥5 reads. We developed an open-source pipeline, Oncoseq, using nf-core standards to streamline data analysis (github.com/chusj-pigu/nf-core-oncoseq). Across 31 samples, AS-WGS achieved a mean on-target coverage of 160X (range: 55–251) and genome-wide coverage of 17X (range: 9–25). Performance was influenced by DNA quality (DIN 8.0) and flow cell characteristics. Among leukemias (n=20), all clinical somatic mutations (47/47, 100%) were detected, including low-VAF (≥5%) variants and large indels such as FLT3 and UBTF internal tandem duplications. VAFs correlated strongly with clinical exome data (r = 0.905). Sixteen of 17 gene fusions (94%) were confidently identified, supported by a mean of 56 reads (range: 24–112). The remaining fusion was supported by only one read, which was attributable to lower quality (DIN 6.9) and tumor purity. AS-WGS detected challenging fusions, such as DUX4::IGH and cytogenetically cryptic NUP98::NSD1. All copy-number variants (CNVs) were accurately identified in 11 of 12 samples, including key alterations like hyperdiploidy. The one missed event was an interstitial chromosome 12 deletion in 25% of cells, at the limit of microarray detection threshold; it was retrospectively visible in AS-WGS data. Focal deletions in CDKN2A/B (n=6), IKZF1 (n=2), and PAX5 (n=1) were all reliably detected. We next assessed detection timing retrospectively using read timestamps. All large CNVs were detectable within the first hour of sequencing. All clonal mutations (VAF 25%) and all detectable fusions reached confident support (≥10 reads) within 28 hours of sequencing, with median 10-reads detection times of 4.2 and 7.4 hours, respectively. AS-WGS signal from PCR-free libraries also supports methylation calling. Using the Marlin classifier (github.com/hovestadt/MARLIN), 7 of 8 B-ALL samples were correctly classified into molecular subgroups: hyperdiploidy, ETV6::RUNX1, Ph+/Ph-like, DUX4-r, PAX5-r, ZNF384-r, and TCF3::PBX1. Notably, classification was confidently achieved within the first 10 minutes of sequencing, before fusion detection. In summary, AS-WGS enables, in a single test, detection of nearly all clinically relevant alterations in pediatric leukemias, including structural variants and mutations, while also supporting methylation-based classification. Optimized protocols and bioinformatics workflows identify most clonal alterations within the first day, with interpretable findings emerging within the first hour. This strategy holds strong potential to transform time-sensitive clinical decision-making and is currently undergoing prospective validation.