8 聚类[Clustering]

8.1 样本-基因 热图[Sample-Gene heatmap]

方法[Method]

在基因水平进行无监督的层次聚类，使用的数据是所有样本基因的 \(log_2(FPKM+1)\) 。 4984 个基因用于聚类， 筛选条件是 \(1<maximum\{log_2(FPKM+1)\}<20\) 并且标准差 \(sd\{log_2(FPKM+1)\} > 0.3\) 。

Unsupervised hierarchical clustering was carried on \(log_2(FPKM+1)\) across samples. 4984 genes used for clustering, which were selected by \(1<maximum\{log_2(FPKM+1)\}<20\) and the standard deviation \(sd\{log_2(FPKM+1)\} > 0.3\).

Figure 8.1: Unsupervised clustering (samples-genes)

PDF 文件 : Heatmap.jpg.

• 要阅读热图中详细的基因名称，请参阅较大的视图 PDF PNG.
• 按层次聚类分成12个组 PDF PNG.
  • 分组信息 Heatmap_cutree12_GroupInfo.xls.
  • 组内信息 Heatmap_cutree12.xls.
• To read the detailed gene names in heatmap, please refer to the larger view PDF PNG.
• To check gene names in sub-groups, a 12-cut view is generated PDF PNG.
  • The sub-group information can be found at Heatmap_cutree12_GroupInfo.xls.
  • The information about genes in each sub-group can be found at Heatmap_cutree12.xls.

8.2 样本-样本 热图[Sample-Sample heatmap]

方法[Method]

在样本水平进行无监督的层次聚类，其中距离定义为 \((1-r^2)\), 其中 \(r\) 是上述所选基因 \(log_2(FPKM+1)\) 的相关系数.

Unsupervised hierarchical clustering was samples, where the distance is defined as \((1-r^2)\), where the \(r\) is the correlation coefficient for \(log_2(FPKM+1)\) on above selected genes.

Figure 8.2: Unsupervised clustering (samples-sample)

PDF 文件 : Heatmap_samplewise.jpg.

8.3 差异表达基因热图[Heatmap only for DEG]

方法[Method]

仅使用 \(FPKM>1\) 的差异表达基因进行监督层次聚类 。

Supervised hierarchical clustering was carried only on DEGs (differentially expresed genes) that \(FPKM>1\).

Figure 8.3: Clustering only for DEGs

PDF 版本 : Heatmap_DEG.jpg.

• 要阅读热图中详细的基因名称，请参阅较大的视图 PDF PNG。

• To read the detailed gene names in heatmap, please refer to the larger view PDF PNG.

---