# The Proteasome-Family-Members-Based Prognostic Model Improves the Risk Classification for Adult Acute Myeloid Leukemia

## TL;DR

This study identifies a three-gene model based on proteasome family members that improves risk prediction and treatment guidance for adult acute myeloid leukemia patients.

## Contribution

A novel three-PSM-based prognostic model is developed to enhance risk classification and therapeutic strategies in AML.

## Key findings

- The model includes PSMB8, PSMG1, and PSMG4, which predict overall survival in AML patients.
- High-risk patients showed worse outcomes and distinct genetic profiles compared to low-risk patients.
- The model improves the ELN risk stratification system and identifies potential therapeutic targets.

## Abstract

Background: The accumulation of diverse molecular and cytogenetic variations contributes to the heterogeneity of acute myeloid leukemia (AML), a cluster of hematologic malignancies that necessitates enhanced risk evaluation for prognostic prediction and therapeutic guidance. The ubiquitin–proteasome system plays a crucial role in AML; however, the specific contributions of 49 core proteasome family members (PSMs) in this context remain largely unexplored. Methods: The expression and survival significance of 49 PSMs in AML were evaluated using the data from BeatAML2.0, TCGA, and the GEO database, mainly through the K-M plots, differential genes enrichment analysis, and candidate compounds screening via R language and statistical software. Results: we employed LASSO and Cox regression analyses and developed a model comprising three PSMs (PSMB8, PSMG1, and PSMG4) aimed at predicting OS in adult AML patients, utilizing expression profiles from the BeatAML2.0 training datasets. Patients with higher risk scores were predominantly found in the AML–M2 subtype, exhibited poorer ELN stratification, showed no complete remission following induction therapies, and had a higher mortality status. Consistently, significantly worse OS was observed in high-risk patients across both the training and three validation datasets, underscoring the robust predictive capability of the three-PSMs model for AML outcomes. This model elucidated the distinct genetic abnormalities landscape between high- and low-risk groups and enhanced the ELN risk stratification system. Ultimately, the three-PSMs risk score captured AML-specific gene expression signatures, providing a molecular basis for selecting potential therapeutic agents. Conclusions: In summary, these findings manifested the significant potential of the PSM model for predicting AML survival and informed treatment strategies.