Delineation of protein networks affected by somatic mutation in neuroblastoma

Aim:The mutational spectrum of neuroblastoma is highly heterogeneous. Here, we aim to delineate protein interaction networks that are recurrently affected by somatic mutations in neuroblastoma using whole exome sequencing.

Method: Using the concepts of network hierarchy and modularity, we developed a novel similarity metric to compare pairs of genes. The metric incorporates information about the structure of the protein interaction network around both compared genes. Combined with unsupervised clustering the metric can identify mutations that group in tightly connected subnetworks.

Results: We identified 257 somatic non-silent mutations in 24 neuroblastoma samples by whole exome sequencing. Application of our method revealed 3 subnetworks enriched in somatic mutations. The largest subnetwork (33 genes of which 20 mutated) was affected in 14 high-risk samples and contained 6 directly connected mutated genes, among which bona fide cancer genes such as TP53 and HRAS.

Conclusion: We have shown that somatic mutations are enriched in particular protein subnetworks in neuroblastoma. Our strategy may help to identify recurrently altered pathways in tumors with heterogeneous mutation profiles.