An Integrated Analysis of Tumor Purity of Common Central Nervous System Tumors in Children Based on Machine Learning Methods

Background: Tumor purity is defined as the proportion of cancer cells in the tumor tissue, and its effects on molecular genetics, the immune microenvironment, and the prognosis of children’s central nervous system (CNS) tumors are under-researched.

Methods: We applied random forest machine learning, the InfiniumPurify algorithm, and the ESTIMATE algorithm to estimate the tumor purity of every child’s CNS tumor sample in several published pediatric CNS tumor sample datasets from Gene Expression Omnibus (GEO), aiming to perform an integrated analysis on the tumor purity of children’s CNS tumors.

Results: Only the purity of CNS tumors in children based on the random forest (RF) machine learning method was normally distributed. In addition, the children’s CNS tumor purity was associated with primary clinical pathological and molecular indicators. Enrichment analysis of biological pathways related to the purity of medulloblastoma (MB) revealed some classical signaling pathways associated with MB biology and development-related pathways. According to the correlation analysis between MB purity and the immune microenvironment, three immune-related genes, namely, CD8A, CXCR2, and TNFRSF14, were negatively related to MB purity. In contrast, no significant correlation was detected between immunotherapy-associated markers, such as PD-1, PD-L1, and CTLA4; most infiltrating immune cells; and MB purity. In the tumor purity–related survival analysis of MB, ependymoma (EPN), and children’s high-grade glioma, we discovered a minor effect of tumor purity on the survival of the aforementioned pediatric patients with CNS tumors.

Conclusion: Our purity pediatric pan-CNS tumor analysis provides a deeper understanding and helps with the clinical management of pediatric CNS tumors.