Predicting leukemic transformation in myelodysplastic syndrome using a transcriptomic signature

Area of Science:

• Genomics and Bioinformatics
• Hematology and Oncology
• Computational Biology

Background:

• Transcriptomic models show promise for predicting leukemic transformation in Myelodysplastic Syndromes (MDS).
• Existing models often lack external validation and rely on specific cell populations (CD34+ sorted cells), limiting clinical applicability.
• There is a need for robust, validated prognostic tools for MDS progression.

Purpose of the Study:

• To identify transformation-associated gene expression patterns in MDS.
• To develop and validate a novel prognostic scoring system for predicting leukemic transformation in MDS patients.
• To assess the clinical feasibility of the new model using unsorted cells.

Main Methods:

• Weighted Gene Co-expression Network Analysis (WGCNA) was used to identify transformation-specific gene modules from transcriptomic data (GSE58831).
• The Least Absolute Shrinkage and Selection Operator (LASSO) algorithm was applied to build a prognostic score (MDS15) using a 15-gene signature.
• The MDS15 model was trained on CD34+ sorted cells (GSE58831) and validated on unsorted cells (GSE15061).

Main Results:

• A transformation-specific gene module ('brown') was identified, enriched in transcription-regulating pathways.
• The MDS15 score, based on a 15-gene signature, demonstrated superior predictive power (AUC, C-index) compared to IPSS/WPSS in both training and independent testing sets.
• The model achieved high predictive accuracy (AUC 0.933, C-index 0.86) on the independent cohort of unsorted cells.

Conclusions:

• Gene co-expression analysis successfully identified a crucial gene module for MDS transformation.
• The novel MDS15 prognostic system offers a validated and potentially more clinically applicable tool for predicting leukemic transformation in MDS.
• The MDS15 score's validation on an independent cohort and unsorted cells supports its robustness and real-world utility.