Differential gene expression and genomic patient stratification following left ventricular assist device support
Burns C. Blaxall, PhD*,
Bryn M. Tschannen-Moran, BSci*,
Carmelo A. Milano, MD* and
Walter J. Koch, PhD*,*
* Department of Surgery, Duke University Medical Center, Durham, North Carolina, USA
View larger version (27K):
[in a new window]
|
Figure 1 Multidimensional scaling (MDS) and similarity matrix/exact test demonstrating clear and significant effect of left ventricular assist device (LVAD) support on gene expression. (A) All gene expression data for pre-LVAD (green) and post-LVAD (red) patients ( 6,800 average difference values from Affymetrix per patient) were subjected to MDS using Pearson dissimilarities. The image represents convergence of the MDS algorithm, stress (measure of image validity) = 0.03. (B) Similarity matrix based on the Pearson correlation of the 12 samples (6 pre- and 6 post-LVAD). Following an exact test (random combinations of the 12 samples assigned to two groups), the most significantly divergent was the actual pre- and post-LVAD grouping, with a significance value of 0.002, illustrating the significant effect of LVAD support on gene expression.
|
|
View larger version (8K):
[in a new window]
|
Figure 2 Multidimensional scaling (MDS) of pre- and post-left ventricular assist device (LVAD) gene expression. (A) Raw data classified according to segregating patient clinical characteristics were subjected to MDS. Pre-1 (yellow) and post-1 (red) represent nonischemic patients, whereas pre-2 (blue) and post-2 (green) represent ischemic patients. (B) Raw gene expression data from the 530 genes determined to be significant by paired t test analysis, classified according to segregating patient clinical characteristics and subjected to MDS. The images represent convergence of the MDS algorithm, stress for both (measure of image validity) <0.05. Note the limitation of program-assigned color schemes.
|
|
View larger version (20K):
[in a new window]
|
Figure 3 Hierarchical clustering using Cluster and Treeview. Data for these programs require different names for each sample, hence 1-3 represent nonischemic patients and 4-6 represent ischemic patients. (A) Hierarchical clustering of raw data; note the segregation of the nonischemic patients and relative similarity of ischemic patients post-left ventricular assist device (LVAD). (B) Hierarchical clustering of raw data from the 530 genes found to be significantly regulated by the paired t test. Note the distinct separation of pre- and post-LVAD, along with segregation of the nonischemic and ischemic patients pre-LVAD.
|
|
View larger version (29K):
[in a new window]
|
Figure 5 Clustering of genes by functional process. Genes found to be significant by the paired t test were clustered according to the biologic process by linking the GenBank accession number with the Gene Ontology database. (A) Small screenshot of the entire cluster by Gene Ontology biologic process. (B) Number and percentage of significant genes in relation to representation on entire chip. Note that not all genes represented on the chip are annotated in the Gene Ontology data base, hence only 397 of the 530 genes are represented here.
|
|
|
-actin, demonstrating the divergent regulation of gene expression following LVAD support in different heart failure etiologies. (B) Signal to glyceraldehyde-6-phosphate dehydrogenase ratio of Northern blot data for ß-MHC and SK