IntroductionAcute lymphoblastic leukemia (ALL) is the most common pediatric malignancy, characterised by the uncontrolled proliferation of lymphoid progenitor cells. The genetic landscape of ALL is highly heterogeneous, encompassing a wide range of structural variants (SVs) and single-nucleotide variants (SNVs) that play a critical role in leukemogenesis. Notably, SVs, including chromosomal rearrangements, deletions, and copy number alterations (CNAs), are enriched in pediatric ALL (pALL) compared to adult cases [1]. These genetic abnormalities define distinct molecular subtypes and are crucial for predicting prognosis and response to therapy [2].The latest update of the World Health Organization (WHO) [3] and International Consensus Classification (ICC) [4] increasingly emphasise the role of molecular alterations in defining leukemia subtypes. This is exemplified by the inclusion, for the first time, of entities defined by SNVs (IKZF1 N159Y, PAX5 P80R). In parallel, modern treatment protocols, such as the ALLTogether (NCT04307576) consortium guidelines, have implemented risk stratification strategies based on fusions and CNA profiling, as well as targeted inhibitors for ABL-class patients and immunotherapy into frontline therapy. These advancements highlight the heterogeneity of diagnostic entities, and thus the need to combine genomic approaches for a precise molecular characterisation that exceeds the capabilities of current standard-of-care (SoC) methods.Traditionally, the SoC methods for genetic diagnosis in ALL relied on conventional cytogenetics, including chromosome banding analysis (CBA) and fluorescence in situ hybridisation (FISH). While these methods have been instrumental in detecting recurrent chromosomal abnormalities, they exhibit significant limitations, such as low resolution, limited detection capacity for cryptic alterations, and reliance on viable metaphases [5]. To overcome these constraints, emerging methodologies such as multiplex ligation-dependent probe amplification (MLPA), targeted next-generation sequencing (t-NGS), digital multiplex ligation-dependent probe amplification (dMLPA), RNA sequencing (RNAseq), or optical genome mapping (OGM) have been progressively implemented into the diagnostic workup, offering enhanced sensitivity and broader detection capabilities [6,7,8].Despite the improvement potential of these novel technologies, each method has inherent strengths and weaknesses in detecting specific types of genetic alterations. To date, no single approach has demonstrated a comprehensive coverage of the entire mutational spectrum of ALL, leaving the optimal diagnostic strategy for clinical practice unresolved [9]. A systematic evaluation of these methodologies is required to determine the most effective approach for an accurate and efficient diagnosis in the clinical setting.In this study, we analysed a cohort of 60 pALL patients using the standard-of-care, OGM, t-NGS, MLPA, dMLPA, and RNAseq (when sufficient genetic material was available). The primary aim was to assess the diagnostic yield of each method, both individually and in combination, to identify the most robust strategy for the comprehensive genomic characterisation of pALL in the clinical setting. To date, this cohort represents the largest pALL series characterised by OGM in a clinical setting within a single institution, providing valuable insights into the optimal diagnostic workflow for this disease.Material and methodsPatients and samplesA total of 60 bone marrow (BM) or peripheral blood (PB) samples (55 diagnoses and 5 relapses) were obtained from pALL patients (49 B-ALL, 11 T-ALL) referred to our institution between August 2021 and August 2024. The cohort included 36 males (60%) and 24 females (40%), with a median age of 5 years (range 1–16). The median blast percentage was 90% (range 20–100). Patients were selected solely based on the availability of high-quality samples to ensure that the cohort accurately reflected real-world clinical testing conditions.Written consents were obtained from parents or legal guardians of all patients according to the recommendations of the Human Rights Declaration and the Helsinki Conference. This study was approved by the institutional ethics committee for clinical research.Standard-of-care baselineImmunophenotyping was performed by flow cytometry following standardised procedures. The antibody panel included anti-CD45, CD34, CD123, CD10, CD19, CD20, CD22, CD9, CD24, CD25, CD15, NG2, CD66c (KORSA), CD33, CD13, cytoplasmic MPO, nuclear TdT, and cytoplasmic CD3 (Beckman Coulter, CA, USA), as well as CD38, cytoplasmic IgM, and kappa/lambda light chains (Dako, CA, USA). Cytogenetic analysis was conducted using G-banding on metaphase chromosomes, with karyotypes interpreted according to the International System for Human Cytogenomic Nomenclature [10]. FISH were performed on interphase nuclei using commercial probes for BCR::ABL1, KMT2A, ETV6::RUNX1, TCF3, CRLF2, ABL2, EPOR, PDGFRB, and JAK2, following the manufacturer’s instructions. Additionally, ETV6::RUNX1 and BCR::ABL1 rearrangements were also assessed by RT-qPCR as described by Gabert et al. [11].DNA and RNA isolationGenomic DNA (gDNA) and total RNA were extracted using the QIAsymphony SP/AS instrument (Qiagen, Valencia, CA) automated platform. DNA extraction was performed with the QIAamp DNA Mini Kit (Qiagen), while RNA was isolated using the RNeasy Midi Kit (Qiagen), following the manufacturer’s instructions. The extracted nucleic acids were quantified using the Qubit Fluorometer (Thermo Fisher Scientific, San Francisco, CA, USA) with the Qubit dsDNA High Sensitivity Assay Kit for DNA and the Qubit RNA HS Assay Kit for RNA.Emerging methodsMultiplex ligation-dependent probe amplificationThe gDNA was isolated from BM or PB samples as described above. MLPA was performed in 100 ng of gDNA using the SALSA MLPA P335 (BTG1, CDKN2A/B, EBF1, ETV6, IKZF1, PAR1 region, PAX5 and RB1) (MRC-Holland, Amsterdam, The Netherlands) following the manufacturer’s instructions. Capillary electrophoresis was carried out on a SeqStudio Genetic Analyzer (Applied Biosystems, Foster City, CA, USA), and data were analysed using Coffalyser.Net software (MRC-Holland) according to established guidelines.Digital multiplex ligation-dependent probe amplificationdMLPA was performed on 50 ng of gDNA using SALSA digitalMLPA D007 Acute Lymphoblastic Leukemia probemix (MRC-Holland) according manufacturer’s recommendations. The probemix includes target probes to identify recurrent microdeletions or amplifications, and karyotyping probes to detect gross chromosomal abnormalities along all chromosomes.Reactions were pooled and sequenced on a MiSeq sequencer with 150 bp single-read chemistry (Illumina, San Diego, CA, USA). Coffalyser digitalMLPA software (MRC-Holland) was used to analyse the copy number status. Regions with a probe ratio value around 1.0 (±0.15) were considered unaffected, while an increased or decreased value indicated the presence of a gain or loss, respectively. Leukemic cell burden (LCB) was considered to interpret the results. Subclonal CNAs were only reported if consecutive probes had dosage values unambiguously falling outside the range but not reaching the expected ratio for a loss/gain based on the LCB, and also compared with other affected regions within the same sample.Optical genome mappingOGM was conducted on fresh (less than 24 h after sample collection) or frozen PB or BM samples according to the standard protocol (Bionano Genomics, San Diego, CA, USA). Briefly, ultra-high molecular weight genomic DNA (UHMW-DNA) was isolated and labelled using DLE-1 enzyme and the Bionano Prep direct labelling and staining (DLS) protocol. A total of 750 ng of labelled UHMW-DNA was loaded on a Saphyr G2.3 chip and run on Bionano’s Saphyr for imaging. Quality criteria were as follows: map rates greater than 60%, molecule N50 values >250 kb (for molecules >150 kb), and effective genome coverage >300×. Genome analysis was performed using the human genome GRCh38 as a reference, and Bionano Access 1.6 and Bionano Solve 3.6 software. Variant calling was performed with Rare Variant Pipeline and Guided assembly with standard filter settings.Next-generation sequencingt-NGS was performed using the ALLseq panel (Gil et al. [12]), designed to detect SNVs, insertions/deletions (indels), CNAs, gene fusions, and gene expression. The full list of targeted genes is provided in Table S1. Briefly, 10 ng of gDNA and RNA were used for library preparation, which was automated on the Ion Chef™ System (Thermo Fisher Scientific). Sequencing was conducted on the Ion S5 sequencer (Thermo Fisher Scientific). Variant calling was performed using the Ion Reporter software (Thermo Fisher Scientific), and variants were considered relevant when their allelic frequency exceeded 3%.RNAseqTotal RNA was extracted from BM or PB as described above. Due to limited sample availability, RNAseq was performed in 20 patients. Quantification and integrity were assessed using a Qubit fluorometer (Thermo Fisher Scientific) and an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA), respectively. RNA libraries were prepared using the TruSeq Stranded Total RNA Library Prep Kit (Illumina), followed by sequencing using paired-end 150 nt reads on the NovaSeq 6000 platform (Illumina), with approximately 20 million reads per sample. Data processing and bioinformatic analysis were conducted using HISAT2 for alignment and STARfusion for fusion detection. As STARfusion often reported multiple fusion genes per sample, including false positives, only high-confidence fusions supported by at least 2 junction reads and 1 spanning fragment read (at least 10,000 nucleotides apart if both genes were on the same chromosome) were considered. Additionally, specific parameters were used to identify IG rearrangements as described by Thomson et al. [13]. Differential gene expression levels were quantified by DESeq2, and diagnostic entities were predicted using the ALLCatchR classifier, with >0.5 score considered relevant.Diagnosis yield assessment across different techniques versus standard-of-care testingThe diagnostic performance of each technique was evaluated both individually and in combination with the other. The diagnostic yield was assessed based on two levels of clinical relevance: (1) Capability to identify entities recognised by the WHO 2022 and/or ICC 2022 classifications, as well as risk stratification markers according to the ALLTogether guidelines; and (2) Any other pathogenic alterations affecting known drivers in pALL or associated with prognosis but not currently incorporated into clinical risk stratification. Diagnostic entities were assigned according to the WHO 2022/ICC classification if a defining driver or a compatible gene expression profile (GEP) was identified by any of the methods used.Results were classified at the patient level into three categories based on concordance: (1) Completely concordant: all detected alterations were identified by the compared techniques; (2) Partially concordant: some but not all abnormalities were detected by both techniques; and (3) Discordant: entirely different abnormalities were identified by the techniques in comparison. Clinically relevant alterations identified exclusively by a single technique were further validated using orthogonal methods.Statistical analysisCategorical variables were compared using Fisher’s exact test or the chi-square test, as appropriate, while continuous variables were analysed using the Mann–Whitney U test or Student’s t-test. The sensitivity and specificity values of each method were calculated against the standard of care. Co-segregation of genetic alterations was analysed using the “somaticInteraction” function from the maftools package. All statistical analyses were performed using R software (version 4.4.2), with a significance threshold set at p