Finding Non-coding Drivers in Childhood Cancer

Abstract

Childhood malignancies are still the leading cause of disease-related deaths in developed countries. Extensive studies are still necessary to understand the carcinogenesis of these tumors, especially concerning the role of non-coding mutations. It was reported that about 80% of the non-coding region in the human genome has biochemical functions. This means that including the non-coding region in childhood cancer driver discovery may help in revealing novel oncogenesis mechanisms. However, finding cancer drivers in this region is difficult due to incomplete annotations and inconsistent mutation rates. Because of this, different driver discovery methods with varying assumptions might not perform equally well in all sections of the non-coding region. To combine the strengths of individual methods, a combination of 3 programs was used to find potential cancer drivers in the non-coding region. DriverPower identified candidate non-coding cancer drivers by modeling global background mutation rates. OncodriveCLUSTL performed local permutations to find mutations that cluster together in a genomic element. Lastly, OncodriveFML searched for mutated regulatory elements with higher-than-expected average functional impact scores. The p-values produced by the 3 programs were combined with Brown’s method. After p-values combination and manual curation, a list of 148 probable non-coding drivers was created. Some of these candidates were observed to be associated with alterations in gene expression. However, the conditions in which gene expression could be altered by the probable non-coding cancer drivers may vary. A single mutation that affected a critical region in the TERT promoter could elevate TERT expression. While in the PGAM1 promoter, interactions between SNVs in this region seemed to be important in altering PGAM1 expression. Lastly, identical SNVs in the B3GALT2 promoter were observed to be associated with varying expression levels in different cancer supergroups. This might imply the importance of interaction between underlying transcription control and the candidate non-coding drivers. These observations and the list of candidate non-coding drivers may serve as guides to find targets for further study to understand non-coding drivers in childhood cancer. Understanding the involvement of non-coding cancer drivers in deregulation of gene expression may provide insights into the progression of childhood cancer.