# A Multi-Omics and machine learning platelet-related prognostic signature in multiple myeloma

## TL;DR

This study explores how platelet-related genes affect multiple myeloma progression and creates a genetic risk model to predict patient outcomes.

## Contribution

A novel platelet-related prognostic signature was developed using multi-omics data and machine learning for multiple myeloma risk stratification.

## Key findings

- Platelet activity promotes tumor cell proliferation and suppresses apoptosis in multiple myeloma.
- A 13-gene platelet-related prognostic signature effectively stratifies MM patients into distinct risk groups.
- The integrated model combining genetic and clinical data shows improved predictive ability for patient outcomes.

## Abstract

While platelets are well-documented contributors to tumorigenesis, their role in multiple myeloma (MM) progression and risk stratification remains underexplored. To assess platelet function in MM and verify the prognostic value of platelet-related genes(PRGs) in patients with multiple myeloma, further providing new ideas for the development of MM. We combined the clinical assessment of platelet activation in MM patients with functional co-culture experiments using MM cell lines (RPMI8226, MM.1 S) to investigate platelet-driven tumor progression. Additionally, an integrated analysis of bulk (GSE124310) and single-cell transcriptomic datasets (GSE6477, TCGA-MM data, GSE4581, GSE24080,and GSE136337) was performed to identify platelet-related prognostic genes (PRGs). Through single-cell RNA sequencing, we identified aberrant erythroid-megakaryocyte components in multiple myeloma and further demonstrated dysfunctional platelet activity that promotes tumor cell proliferation and suppresses apoptosis. Using comprehensive bioinformatic screening across 116 algorithms, we identified a combined forward stepwise Cox and Ridge regression model as optimal and established a 13-gene platelet-related prognostic signature. The genetic risk model effectively stratified MM patients into distinct prognostic groups, with high-risk patients exhibiting poorer outcomes in both training and validation cohorts. Finally, we integrated the genetic risk model and clinically relevant information and visualized it with dynamic Nomogram plots, and the ROC and DCA curves demonstrated that the integrated model had better predictive ability. Our study establishes a significant association between platelet activity and disease progression in MM. The platelet-related prognostic signature we developed is correlated with patient outcomes and may have utility in risk stratification.

The online version contains supplementary material available at 10.1007/s00277-026-06867-8.