A computational approach for the discovery of significant cancer genes by weighted mutation and asymmetric spreading strength in networks

Abstract

Identifying significantly mutated genes in cancer is essential for understanding the mechanisms of tumor initiation and progression. This task is a key challenge since large-scale genomic studies have reported an endless number of genes mutated at a shallow frequency. Towards uncovering infrequently mutated genes, gene interaction networks combined with mutation data have been explored. This work proposes Discovering Significant Cancer Genes (DiSCaGe), a computational method for discovering significant genes for cancer. DiSCaGe computes a mutation score for the genes based on the type of mutations they have. The influence received for their neighbors in the network is also considered and obtained through an asymmetric spreading strength applied to a consensus gene network. DiSCaGe produces a ranking of prioritized possible cancer genes. An experimental evaluation with six types of cancer revealed the potential of DiSCaGe for discovering known and possible novel significant cancer genes.

The central role of cancer driver genes in pathway networks according to persistence homology

“…Considering traditional measures of complex networks, many drivers are hubs [29], although there are drivers with a low degree [11]. In fact, one of the challenges is finding drivers in the long tail of the distributions associated with PPIN and mutation data [8] since most methods suffer from "ascertainment bias", favouring frequently mutated genes and hubs [30]. Fig 8 shows that the impact on the reduction of H2 structures is not fully correlated with a high degree, indicating that the genes in Fig 7 have a topological role associated with high-dimensional structures, which go beyond traditional centrality measures that only consider edges.…”

Abordagens computacionais para a descoberta de genes significativos para o câncer

“…Como decorrência da pesquisa de doutorado, vários trabalhos de pesquisa foram submetidos e publicados em conferências e revistas: 1) Um resumo expandido no Simpósio Brasileiro de Computac ¸ão Aplicada a Saúde de 2019 [Cutigi et al 2019b]; 2) Um artigo completo no Brazilian Symposium on Bioinformatics de 2019 [Cutigi et al 2019a]; 3) Um artigo completo, como segundo autor, no Simpósio Brasileiro de Computac ¸ão Aplicada a Saúde de 2020 [Ramos et al 2020]; 4) Um artigo completo na revista Journal of Bioinformatics and Computational Biology em 2020 [Cutigi et al 2020a]; 5) Um artigo completo no Brazilian Symposium on Bioinformatics de 2021 [Cutigi et al 2020b]; e 6) Um artigo completo na revista Nature Scientific Reports em 2021 [Cutigi et al 2021].…”

Identifying potential risk genes for clear cell renal cell carcinoma with deep reinforcement learning

“…Muffinn [28] identified cancer risk genes through network propagation, taking into account mutations not only in individual genes but also in their neighbors within the protein-protein interaction (PPI) network. DiSCaGe [29] calculated a gene mutation score using an asymmetric spreading strength based on the type of mutations and the PPI network, then produced a ranking of prioritized cancer risk genes. HotNet2 [29] used an insulated heat diffusion process to identify can-cer risk genes by propagating heat through the PPI network.…”

“…DiSCaGe [29] calculated a gene mutation score using an asymmetric spreading strength based on the type of mutations and the PPI network, then produced a ranking of prioritized cancer risk genes. HotNet2 [29] used an insulated heat diffusion process to identify can-cer risk genes by propagating heat through the PPI network. nCOP [30] employed a heuristic search method to select connected subnetworks from the PPI network based on the mutation data of cancer patients, and then ranked cancer risk genes based on the frequencies of genes appearing in these subnetworks.…”