Extracting Common Signatures To Classify Human Nervous System Cancer Data with Bioinformatics Approaches

Abstract

Abstract Cancer is a major life-threatening issue in Sri Lanka as well as worldwide. Meta-analysis of Human Nervous System (HNS) cancers is a gap in current cancer research. The proposed research is focused on identifying common and unique signatures (patterns) in HNS cancers using bioinformatics approaches. We have used fifteen mutation cancer datasets and four gene expression datasets downloaded from the cBio Portal and analyzed the data comprehensively. In mutation analysis, single nucleotide polymorphisms (SNP) were used. We have used clustering approaches, association rules and community detection methods for filtering biologically meaningful gene clusters. Gene expression data were also analyzed using these methods. We have validated the identified gene clusters with the Mentha Human Interactome using the Reactome pathways and silhouette statistics. The UniProt, Ensemble, PDB and Mentha databases were used for data annotation and validation. We found 5 association rules, 19 MCODE gene clusters, 16 ClusterOne gene clusters and 622 M-Cliques from the HNS cancer profiles. These 669 gene clusters were identified as Human Nervous system cancer (HNS) gene combinations. These signatures and their genes and proteins are visualized as networks. The genetic, protein level information of each gene and actionable drug details for treatment are annotated and visually presented. We have created the HNS interactome based on the network results. There are 569 genes (nodes) and 7282 edges (interactions). Moreover, we have suggested 8 gene panels for HNS cancer detection. Furthermore, we have created an information enriched dataset for HNS cancer mutations. We conclude that there are highly connected sub-networks in HNS cancer. Since tumours have high cross cancer similarity and heterogeneity it is difficult to find a pattern for mutational level and gene expression level only using computational methods. Our methodology is developed as a reusable component which can be used to analyze different types of cancers. Finally, gene signatures can be practically realized as gene panels to test for HNS cancers.