A prognostic survival model based on metabolism-related gene expression in plasma cell myeloma

Han-ying Huang,Yun Wang,Wei-da Wang,Xiao-li Wei,Robert Peter Gale,Jin-yuan Li,Qian-yi Zhang,Ling-ling Shu,Liang Li,Juan Li,Huan-xin Lin,Yang Liang

DOI: https://doi.org/10.1038/s41375-021-01206-4

2021-03-08

Leukemia

Abstract:Accurate survival prediction of persons with plasma cell myeloma (PCM) is challenging. We interrogated clinical and laboratory co-variates and RNA matrices of 1040 subjects with PCM from public datasets in the Gene Expression Omnibus database in training (<i>N</i> = 1) and validation (<i>N</i> = 2) datasets. Genes regulating plasma cell metabolism correlated with survival were identified and seven used to build a metabolic risk score using Lasso Cox regression analyses. The score had robust predictive performance with 5-year survival area under the curve (AUCs): 0.71 (95% confidence interval, 0.65, 0.76), 0.88 (0.67, 1.00) and 0.64 (0.57, 0.70). Subjects in the high‐risk training cohort (score &gt; median) had worse 5-year survival compared with those in the low‐risk cohort (62% [55, 68%] vs. 85% [80, 90%]; <i>p</i> &lt; 0.001). This was also so for the validation cohorts. A nomogram combining metabolic risk score with Revised International Staging System (R-ISS) score increased survival prediction from an AUC = 0.63 [0.58, 0.69] to an AUC = 0.73 [0.66, 0.78]; <i>p</i> = 0.015. Modelling predictions were confirmed in in vitro tests with PCM cell lines. Our metabolic risk score increases survival prediction accuracy in PCM.

oncology,hematology