Clustering of RPPA data: consensus hierarchical
Head and Neck Squamous Cell Carcinoma (Primary solid tumor)
16 April 2014  |  analyses__2014_04_16
Maintainer Information
Citation Information
Maintained by TCGA GDAC Team (Broad Institute/MD Anderson Cancer Center/Harvard Medical School)
Cite as Broad Institute TCGA Genome Data Analysis Center (2014): Clustering of RPPA data: consensus hierarchical. Broad Institute of MIT and Harvard. doi:10.7908/C1416VNS
Overview
Introduction

This pipeline calculates clusters based on consensus hierarchical clustering with agglomerative average linkage , . This pipeline has the following features:

  1. Classify samples into consensus clusters.

  2. Determine differentially expressed marker proteins for each subtype.

Summary

160 proteins were selected. Consensus average linkage hierarchical clustering of 212 samples and 160 proteins identified 3 subtypes with the stability of the clustering increasing for k = 2 to k = 8 and the average silhouette width calculation for selecting the robust clusters.

Results
Protein expression patterns of molecular subtypes

Figure 1.  Get High-res Image Samples were separated into 3 clusters. Shown are 212 samples and 160 marker proteins. The color bar of the row indicates the marker proteins for the corresponding cluster.

Figure 2.  Get High-res Image Heatmap with a standard hierarchical clustering for 212 samples and 160 proteins.

Consensus and correlation matrix

Figure 3.  Get High-res Image The consensus matrix after clustering shows 3 clusters with limited overlap between clusters.

Figure 4.  Get High-res Image The correlation matrix also shows 3 clusters.

Silhouette widths

Figure 5.  Get High-res Image The silhouette width was calculated for each sample and each value of k. The left panel shows the average silhouette width across all samples for each tested k (left panel). The right panels shows assignments of clusters to samples and the silhouette width of each sample for the most robust clustering.

Samples assignment with silhouette width

Table 1.  Get Full Table List of samples with 3 subtypes and silhouette width.

SampleName cluster silhouetteValue
TCGA-BA-5152-01A-11-2072-20 1 0.15
TCGA-BA-5559-01A-21-2072-20 1 0.32
TCGA-BA-6871-01A-13-2073-20 1 0.22
TCGA-CN-5361-01A-21-2072-20 1 0.17
TCGA-CN-5366-01A-21-2072-20 1 -0.11
TCGA-CN-5374-01A-21-2072-20 1 0.19
TCGA-CN-6013-01A-13-2072-20 1 0.23
TCGA-CN-6021-01A-13-2073-20 1 0.083
TCGA-CN-6023-01A-13-2073-20 1 0.26
TCGA-CN-6989-01A-13-2073-20 1 0.0075

Table 2.  Get Full Table List of samples belonging to each cluster in different k clusters.

TCGA_ID K.2 K.3 K.4 K.5 K.6 K.7 K.8
TCGA-BA-4078-01A-21-2072-20 1 3 3 1 4 5 5
TCGA-BA-5152-01A-11-2072-20 1 1 2 2 3 3 8
TCGA-BA-5153-01A-21-2072-20 1 3 3 1 4 5 5
TCGA-BA-5555-01A-21-2072-20 1 3 3 1 4 5 5
TCGA-BA-5559-01A-21-2072-20 1 1 2 2 3 3 8
TCGA-BA-6869-01A-13-2073-20 1 3 3 1 4 5 5
TCGA-BA-6871-01A-13-2073-20 1 1 2 2 3 3 8
TCGA-BA-6872-01A-13-2073-20 1 3 3 5 1 1 1
TCGA-CN-4722-01A-21-2072-20 1 3 3 1 4 5 5
TCGA-CN-4725-01A-21-2072-20 1 3 1 4 5 7 7
Marker proteins of each subtype

Samples most representative of the clusters, hereby called core samples were identified based on positive silhouette width, indicating higher similarity to their own class than to any other class member. Core samples were used to select differentially expressed marker proteins for each subtype by comparing the subclass versus the other subclasses, using Student's t-test.

Table 3.  Get Full Table List of marker proteins with p <= 0.05 (The positive value of column difference means protein is upregulated in this subtype and vice versa).

Composite.Element.REF p difference q subclass
CLDN7|CLAUDIN-7-R-V 0.00066 1.4 0.0053 1
CDH1|E-CADHERIN-R-V 0.000036 0.87 0.00075 1
BCL2L11|BIM-R-V 0.000016 0.85 0.00062 1
BCL2|BCL-2-M-V 0.0043 0.78 0.017 1
CASP7|CASPASE-7_CLEAVEDD198-R-C 0.0087 0.76 0.031 1
ATM|ATM-R-C 0.00061 0.66 0.0053 1
CCNB1|CYCLIN_B1-R-V 0.000091 0.6 0.0012 1
CTNNB1|BETA-CATENIN-R-V 0.0033 0.57 0.014 1
LCK|LCK-R-V 0.0015 0.54 0.0092 1
SYK|SYK-M-V 0.0028 0.53 0.013 1
Methods & Data
Input

The RPPA Level 3 data was used as the input for clustering; protein measurements corrected by median centering across antibodies.

Consensus Hierarchical Clustering

Consensus Hierarchical clustering is a resampling-based clustering. It provides for a method to represent the consensus across multiple runs of a clustering algorithm and to assess the stability of the discovered clusters. To this end, perturbations of the original data are simulated by resampling techniques , .

Silhouette Width

Silhouette width is defined as the ratio of average distance of each sample to samples in the same cluster to the smallest distance to samples not in the same cluster. If silhouette width is close to 1, it means that sample is well clustered. If silhouette width is close to -1, it means that sample is misclassified .

Download Results

In addition to the links below, the full results of the analysis summarized in this report can also be downloaded programmatically using firehose_get, or interactively from either the Broad GDAC website or TCGA Data Coordination Center Portal.

References
[1] Bmonti, S., Tamayo, P., Mesirov, J. & Golub, T.R., Consensus Clustering: A Resampling-Based Method for Class Discovery and Visualization of protein Expression Microarray Data, Machine Learning:91-118 (2003)
[3] Rousseeuw, P.J., Silhouettes: A graphical aid to the interpretation and validation of cluster analysis., J. Comput. Appl. Math. 20:53-65 (1987)