Bioinformatics analyses of retinoblastoma reveal the retinoblastoma progression subtypes.

INTRODUCTION: Retinoblastoma (RB) is one common pediatric malignant tumor with dismal outcomes. Heterogeneity of RB and subtypes of RB were identified but the association between the subtypes of RB and RB progression have not been fully investigated.
METHODS: Four public datasets were downloaded from Gene expression omnibus and normalization was performed to remove batch effect. Two public datasets were explored to obtain the RB progression gene signatures by differentially expression analysis while another two datasets were iterated for RB subtypes identification using consensus clustering. After the RB progressive subtype gene signatures were identified, we tested the diagnostic capacity of these gene signatures by receiver operation curve.
RESULTS: Three hundreds and forty six genes that were enriched in cell cycle were identified as the progression signature in RB from two independent datasets. Four subtypes of RB were stratified by consensus clustering. A total of 21 genes from RB progression signature were differentially expressed between RB subtypes. One subtype with low expression cell division genes have less progression of all four subtypes. A panel of five RB subtype genes (CLUL1, CNGB1, ROM1, LRRC39 and RDH12) predict progression of RB.
CONCLUSION: Retinoblastoma is a highly heterogeneous tumor and the level of cell cycle related gene expression is associated with RB progression. A subpopulation of RB with high expression of visual perception has less progressive features. LRRC39 is potentially the RB progression subtype biomarker.