Table 1 Three NA method groups based on how input data are processed

Within-network-only

Given two PPI networks, each node’s feature is calculated using only the topological information within the given node’s own network, hence the group name. The nodes’ topological features, which summarize the nodes’ extended PPI network neighborhoods, are then used in various alignment processes (section “Methods – Description of existing NA methods”). For state-of-the-art NA methods from this group, the topological features are based on graphlets [17], which are subgraphs, i.e., small building blocks of networks.

Isolated-within-and-across-network

Given two PPI networks and also sequence information for nodes across networks, each node’s topological feature is calculated in the same way as by within-network-only methods, and only afterwards is the sequence information combined with the topological features. The group name comes from the fact that both within-network topological and across-network sequence information are used, but the two are initially processed in isolation from each other and are combined only after the fact. Then, the combined data are used in various alignment processes (section “Methods – Description of existing NA methods”). Note that within-network-only methods can easily be used as isolated-within-and-across-network methods when sequence information is available; the latter lead to better alignments than the former [11].

Integrated-within-and-across-network

Given two PPI networks and sequence information for nodes across networks, the two networks are first “integrated” into one by adding across-network “anchor” links (edges) between the highly sequence-similar proteins and only then is any feature extraction or alignment done. So, the third group uses both within-network topological and across-network sequence information. But, they first integrate the two data types and only then process them, hence the group name.