基于聚类的癌症分子网络构建与致病基因分析

Construction of Cancer Molecular Network Based on Clustering and Analysis of Pathogenic Genes

疾病基因预测是生物信息学中一个重要的分支,对疾病生物标记进行研究有助于揭开潜在疾病的发病机制,并引导个性化治疗。文章提出一种以已知疾病基因集合引导的疾病网络的构建方法,并从这个疾病网络预测疾病相关基因。先采用Diffusion Kernel(扩散核)思想构建6种癌症(肺癌、前列腺癌、乳腺癌、膀胱癌、大肠癌及子宫癌)的疾病分子网络。再用一个复杂网络中的快速社团发现算法在疾病网络上进行聚类,对疾病网络中除已知疾病基因以外的基因打分,从中筛选癌症相关基因并在微阵列数据上验证分类性能。实验结果表明,文章提出的方法能比较有效地构建疾病分子网络,预测与特定癌症相关程度高的基因。

The prediction of disease genes is an important branch in bioinformatics.Researches on disease biomarkers will help uncover the underlying disease pathogenesis and guide personalized treatment.This paper proposes a method to construct disease networks guided by known disease-related genes and predict disease genes.Firstly,diffusion kernel is used to construct six disease networks for six kinds of cancer(lung cancer,prostate cancer,breast cancer,bladder cancer,colorectal cancer,endometrial cancer).Then,a fast clustering algorithm is applied in disease networks after which a novel scoring method is proposed to score candidate genes.The Experimental results show that the proposed method can effectively construct disease molecular networks and predict the genes that are highly associated with specific cancers.