Expansion of datasets for functional enrichment analyses
by clustering
As outlined above, our example dataset generated a relatively small
number of significantly regulated proteins, which limits the power of
functional enrichment analyses. To expand the set of proteins for such
analyses we performed non-biased k-means clustering, which resulted in
effectively segregating 37 of the total 42 significantly regulated
proteins into one of two clusters, which essentially correspond to
up-regulated (cluster 6) and down-regulated (cluster 1) groups of
proteins. Cluster 1 was larger and more inclusive, including 20 of the
21 down-regulated proteins and 185 non-significant proteins showing the
same overall abundance pattern. Cluster 6 contained 17 of 21
up-regulated proteins and 37 non-significant proteins with a similar
overall abundance pattern (Figure 4).