@juda
2016-07-31T02:41:41.000000Z
字数 3665
阅读 1388
gene gene network: Since most of genes don't interact with one another, except gene 7316. G7316 is a ubiquitin C, thus it interact with so many other protein. That explains most of genes have small degree centrality while a bulge appears around numeric 0.55 on gene-gene curve.
disease disease network: Because most of diseases have similar side effect, the density of the disease disease network is higher than other network. Consequently, the curve is smoother than other curves and like a right triangle.
drug drug network: The drug drug network is more dense than gene gene network but less dense than disease disease network. Compared with gene gene network, many drugs are similar to plentiful other drugs. Thus the curve is like a right trapezoid.
The degree centrality for a node is the fraction of nodes it is connected to.
gene gene network: It's interesting that it exists two crests. When removing the G7316 from gene gene network, we get only one crest(ignore zero value), which imples that G7316 divides the network roughly into 2 parts and the size of two parts are unbalanced, thus the entire chart exists two crests.
disease disease network: Since the higher density of the network, a disease is easier to contact other diseases. Therefore, the closeness centrality of most diseases is more than 0.4.
drug drug network: The curve represents the density of the network and shows that drug drug network is least dense.
Closeness centrality of a node u is the reciprocal of the sum of the shortest path distances from u to all n−1 other nodes.
Both four plots show that the density is low. If zooming in the phenotype drug network curve, the shape of the curve is like other three curve. As for gene disease network, due to the fact that the amount of edges is small, the variance is high and the fitted curve fluctuate strongly.
For the gene disease network, the crest is close to 0, which reflects the gene disease network is sparse. The disease phenotype network and the phenotype drug network show the similar density. The shape of drug gene network curve shows that the density of the network isn't even. In other words, its shape of network is like a stick.
The closeness of a node is the distance to all other nodes in the graph or in the case that the graph is not connected to all other nodes in the connected component containing that node.
When deleting disease disease network, drug gene network, gene disease network and phenotype drug network, the statistics are similar. The obvious difference is that the lower extreme is 0, which consistent with the degree distribution of these network -- the crest is close to 0.
When deleting disease phenotype network, the lower quartil decrease, because the disease phenotype network has the most edges. Even though disease disease network also has so many edges, its degree distribution is smoother, which is the reason why the low quartil doesn't change obviously.
When deleting drug drug network, the lower extreme doesn't change, which is because all recorded drug can either interact with phenotype or gene.
When deleting gene gene network, both the distribution and median change obviously. The reason is that the number of edges of gene gene network is a little large and most of the degree of each nodes are very small.
The plot show that when we delete the gene disease network or drug gene network, the fusing graph hardly changes. When we delete the disease phenotype network, phenotype drug network or drug drug network, the fusing graph changes slightly. However, when we delete the gene gene network or disease disease network, the fusing network changes a lot. Actually, the shape of gene gene network and disease disease network are much different from other network. The former is like a funnel and the latter is like a spiked ball.
Transitivity measures how many possible triangles are identified by the number of “triads” (two edges with a shared vertex).
Assortativity measures the similarity of connections in the graph with respect to the node degree.