Xi Jianing, Wang Minghui, Li Ao
School of Information Science and Technology, University of Science and Technology of China, Hefei AH 230027, People's Republic of China.
Mol Biosyst. 2017 Sep 26;13(10):2135-2144. doi: 10.1039/c7mb00303j.
The accumulating availability of next-generation sequencing data offers an opportunity to pinpoint driver genes that are causally implicated in oncogenesis through computational models. Despite previous efforts made regarding this challenging problem, there is still room for improvement in the driver gene identification accuracy. In this paper, we propose a novel integrated approach called IntDriver for prioritizing driver genes. Based on a matrix factorization framework, IntDriver can effectively incorporate functional information from both the interaction network and Gene Ontology similarity, and detect driver genes mutated in different sets of patients at the same time. When evaluated through known benchmarking driver genes, the top ranked genes of our result show highly significant enrichment for the known genes. Meanwhile, IntDriver also detects some known driver genes that are not found by the other competing approaches. When measured by precision, recall and F1 score, the performances of our approach are comparable or increased in comparison to the competing approaches.
下一代测序数据的不断积累，为通过计算模型精确找出在肿瘤发生过程中有因果关系的驱动基因提供了契机。尽管此前针对这个具有挑战性的问题已经做了诸多努力，但在驱动基因识别准确性方面仍有提升空间。在本文中，我们提出了一种名为IntDriver的新型综合方法，用于对驱动基因进行优先级排序。基于矩阵分解框架，IntDriver能够有效地整合来自相互作用网络和基因本体相似性的功能信息，并同时检测在不同患者组中发生突变的驱动基因。通过已知的基准驱动基因进行评估时，我们结果中排名靠前的基因显示出对已知基因的高度显著富集。同时，IntDriver还检测到了一些其他竞争方法未发现的已知驱动基因。当通过精确率、召回率和F1分数来衡量时，我们方法的性能与竞争方法相比具有可比性或有所提高。
