B-lineage acute lymphoblastic leukemia (B-ALL) is classified into more than 20 molecular subtypes, and next-generation sequencing has facilitated the identification of these with high sensitivity. Bulk RNA-seq analysis of bone marrow was realized to identify molecular subtypes in Mexican pediatric patients with B-ALL. High hyperdiploidy (27.3%) was the most frequent molecular subtype, followed by DUX4 (13.6%), TCF3::PBX1 (9.1%), ETV6::RUNX1 (9.1%), Ph-like (9.1%), ETV6::RUNX1-like (9.1%), PAX5alt (4.5%), Ph (4.5%), KMT2A (4.5%), and ZNF384 (4.5%), with one patient presenting both the PAX5alt and low hypodiploidy subtypes (4.5%). The genes TYK2, SEMA6A, FLT3, NRAS, SETD2, JAK2, NT5C2, RAG1, and SPATS2L harbor deleterious missense variants across different B-ALL molecular subtypes. The Ph-like subtype exhibited mutations in STAT2, ADGRF1, TCF3, BCR, JAK2, and NRAS with overexpression of the CRLF2 gene. The DUX4 subtype showed mutually exclusive missense variants in the PDGRFA gene. Here, we have demonstrated the importance of using RNA-seq to facilitate the differential diagnosis of B-ALL with successful detection of gene fusions and mutations. This will aid both patient risk stratification and precision medicine.

Mexico has one of the highest incidences of childhood acute lymphoblastic leukemia worldwide and also reports one of the lowest 5-year overall survival rates; one cause is elevated treatment-related mortality from severe hematologic toxicity, a critical side effect during chemotherapy. Although association studies have identified common genetic variants linked to toxicity, the Mexican population remains incompletely explored. We performed targeted next-generation sequencing of genes involved in methotrexate and mercaptopurine metabolism in 96 Mexican children with ALL, stratified by hematologic toxicity severity according to CTCAE v5.0. Logistic regression revealed nominally significant associations between the TSER*3/TSER*2 genotype (rs45445694) in the 5′ UTR of TYMS severe hematologic toxicity (overdominant model: OR = 3.21; 95% CI = 1.23–8.38; p = 0.014 and codominant model: OR = 3.31; 95% CI = 1.17–9.34; p = 0.047). Additionally, the CCND1 LD block (rs678653/rs7177) also showed nominal associations (overdominant model: OR = 2.60; 95% CI = 1.01–6.71; p = 0.044). None of these associations remained statistically significant after Bonferroni correction for multiple testing (corrected α = 0.000588). Patients with severe toxicity had lower 5-year overall survival than those without severe toxicity (70.2% vs. 90.5%; p = 0.0609), although the difference was not statistically significant. Variants in 5′ and 3′ UTRs regions were markedly enriched compared to coding sequences, underscoring the importance of regulatory region analysis.