Software
|
Clustering
|
Number
|
Number
|
BHI (BP)
|
BHI (CC)
|
BHI (MF)
|
BHI (all)
|
---|
| |
of DE genes
|
of clusters
| | | | |
---|
DGEclust
| Hierarchical | 2,177 | 1 |
0.07
| 0.08 |
0.08
|
0.21
|
| Hierarchical ∗
| | 17 | 0.05 |
0.09
| 0.07 | 0.20 |
|
k-means | | 32 | 0.05 | 0.07 | 0.08 | 0.20 |
DESeq2
| Hierarchical | 7,109 | 1 | 0.06 | 0.08 | 0.08 | 0.20 |
|
k-means | | 59 | 0.06 | 0.08 | 0.07 | 0.21 |
edgeR
| Hierarchical | 5,705 | 1 | 0.06 | 0.08 | 0.08 | 0.20 |
|
k-means | | 53 | 0.06 | 0.07 | 0.08 | 0.21 |
- We computed the BHI scores for each GO domain (biological process, molecular function and cellular component), as well as an overall score. k-means and hierarchical clustering were applied to the regularised log-transformed counts for all genes that were called DE between at least one pair of brain regions by each of the three examined methods, i.e. DGEclust, DESeq2 and edgeR. For k-means, we used an optimal number of clusters equal to \(\sqrt {N_{\textit {DE}}/2}\), where N
DE
is the number of DE genes. For the hierarchical clustering, we used average linkage and a Euclidean distance metric with a cutoff distance of 0.5 to obtain an optimal clustering. For DGEclust, we also applied hierarchical clustering using an internally computed similarity matrix. This is indicated with an asterisk (∗). The highest score in each GO domain is indicated in bold. BHI, biological homogeneity index; BP, biological process; CC, cellular component; DE, differentially expressed; GO, gene ontology; MF, molecular function.