Wang Limei, Xie Weixin, Li Kongning, Wang Zhenzhen, Li Xia, Feng Weixing, Li Jin
College of Intelligent Systems Science and Engineering, Harbin Engineering University, Harbin, China.
Key Laboratory of Tropical Translational Medicine, Ministry of Education, College of Biomedical Information and Engineering, Hainan Medical University, Haikou, China.
Front Genet. 2021 Jul 5;12:647653. doi: 10.3389/fgene.2021.647653. eCollection 2021.
Differential co-expression-based pathway analysis is still limited and not widely used. In most current methods, the pathways were considered as gene sets, but the gene regulation relationships were not considered, and the computational speed was slow. In this article, we proposed a novel Dysregulated Pathway Identification Analysis (DysPIA) method to overcome these shortcomings. We adopted the idea of Correlation by Individual Level Product into analysis and performed a fast enrichment analysis. We constructed a combined gene-pair background which was much more sufficient than the background used in Edge Set Enrichment Analysis. In simulation study, DysPIA was able to identify the causal pathways with high AUC (0.9584 to 0.9896). In p53 mutation data, DysPIA obtained better performance than other methods. It obtained more potential dysregulated pathways that could be literature verified, and it ran much faster (∼1,700-8,000 times faster than other methods when 10,000 permutations). DysPIA was also applied to breast cancer relapse dataset and breast cancer subtype dataset. The results show that DysPIA is effective and has a great biological significance. R packages "DysPIA" and "DysPIAData" are constructed and freely available on R CRAN (https://cran.r-project.org/web/packages/DysPIA/index.html and https://cran.r-project.org/web/packages/DysPIAData/index.html), and on GitHub (https://github.com/lemonwang2020).
基于差异共表达的通路分析仍然有限，尚未得到广泛应用。在目前的大多数方法中，通路被视为基因集，但未考虑基因调控关系，且计算速度较慢。在本文中，我们提出了一种新颖的失调通路识别分析（DysPIA）方法来克服这些缺点。我们将个体水平乘积相关性的思想应用于分析，并进行了快速富集分析。我们构建了一个组合基因对背景，它比边集富集分析中使用的背景更加充分。在模拟研究中，DysPIA能够以高AUC（0.9584至0.9896）识别因果通路。在p53突变数据中，DysPIA比其他方法表现更好。它获得了更多可经文献验证的潜在失调通路，并且运行速度更快（当进行10000次排列时，比其他方法快约1700 - 8000倍）。DysPIA还应用于乳腺癌复发数据集和乳腺癌亚型数据集。结果表明，DysPIA是有效的，具有重要的生物学意义。R包“DysPIA”和“DysPIAData”已构建，并可在R CRAN（https://cran.r-project.org/web/packages/DysPIA/index.html和https://cran.r-project.org/web/packages/DysPIAData/index.html）以及GitHub（https://github.com/lemonwang2020）上免费获取。
