CLINICAL SIGNIFICANCE OF NOVEL BIOMARKERS IN THE EARLY DIAGNOSIS AND PROGNOSIS OF ACUTE MYELOID LEUKEMIA

Abstract

Acute myeloid leukemia (AML) is a heterogeneous hematologic malignancy with genomic alterations, which differ and affect the prognosis and response to treatment. One of the most important issues is the dilemma on whether to include the whole-molecule profiling in the clinical decision-making.  It was a retrospective cohort study where 312 patients with newly-diagnosed de novo AML were involved and received a comprehensive genomic profile in the form of next-generation sequencing panel of 54 genes. The dynamics of clonal architecture and the methodologies of detection of the remaining disease as well as patterns of co-mutation were studied. Multivariate Cox regression, machine learning algorithms and graph neural networks were used to create prognostic models. The mutation was also recurrent with highest percentage of 94.6 as the highest percentages of NPM1, DNMT3A and FLT3 which had highest percentages of 28.5, 26.9 and 24.4 respectively. Molecular risk stratification (54 gene) was much better compared to ELN 2022, and had a very-high-risk subgroup with a median overall survival of 8.4 months. TP53 (HR 3.24), ASXL1 (HR 2.18) and RUNX1 (HR 1.96) are reported to be the most bad prognostic markers and CEBPA bZIP (HR 0.52) and NPM1 (HR 0.58) have good results. Co-mutation analysis showed that there were significant pairwise associations, such as that between NPM1 and FLT3-ITD (OR 3.42) and that TP53 and NPM1 mutually exclude each other (OR 0.12). Longitudinal clonal analysis revealed that CHIP-related mutations continued to exist with treatment and the diversity of the clones at relapse with a significant increase. Sensitivity (10 -6) and concordance (89.4): NGS with error-corrected technique demonstrated a good sensitivity and a good concordance. The largest predictive accuracy of treatment response (AUROC 0.912) was achieved with the use of graph neural networks with mutational networks. The broad-based genomic investigation, which covers mutational, clonal and co-mutation can greatly contribute to the prognostic stratification and therapy of AML. These findings support the idea that the multi-gene NGS panel and analysis of the clonal architecture should be implemented into the routine to simplify the precision medicine plans.