Deconvolution of Bulk Gene Expression Profiles with Single-Cell Transcriptomics to Develop a Cell Type Composition-Based Prognostic Model for Acute Myeloid Leukemia

Acute myeloid leukemia (AML) is one of the malignant hematologic cancers with rapid progress and poor prognosis. Most AML prognostic stratifications focused on the genetic abnormalities. However, none of them were established based on the cell type compositions (CTCs) of peripheral blood or bone marrow aspirates from patients at diagnosis. Here, we sought to develop a novel prognostic model for AML in adults based on the CTCs. First, we applied the CIBERSORT algorithm to estimate the CTCs for patients from two public datasets (GSE6891 and TCGA-LAML) using a custom gene expression signature reference constructed by an AML single-cell RNA sequencing dataset (GSE116256). Then, a CTC-based prognostic model was established using Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression, termed CTC score. The constructed prognostic model CTC score comprised three cell types, GMP-like, HSC-like, and T. Compared with the low-CTC score group, the high-CTC score group showed a 1.57-fold (95% confidence interval (CI), 1.23 to 2.00; P=0.0002) and a 2.32-fold (95% CI, 1.53 to 3.51; P<0.0001) higher overall mortality risk in the training set (GSE6891) and validation set (TCGA-LAML), respectively. When adjusting for age at diagnosis, cytogenetic risk, and karyotype, CTC score remained statistically significant in both the training set (Hazard ratio (HR)=2.25; 95% CI, 1.20 to 4.24; P=0.0119) and the validation set (HR=7.97; 95% CI, 2.95 to 21.56; P<0.0001). We further compared the performance of CTC score with two gene expression-based prognostic scores, the 17-gene leukemic stem cell score (LSC17 score) and the AML prognostic score (APS). It turned out that CTC score achieved comparable performance at 1-year, 2-year, 3-year, and 5-year timepoints and provided independent and additional prognostic information different from the LSC17 score and APS. In conclusion, the CTC score could serve as a powerful prognostic marker for AML and has great potential to conduce clinicians to formulate individualized treatment plans.