Incremental network querying in biological networks

Comparative network analysis finds similarities between a query and a target network by aligning them. Biological network evolves over time; and so does the topology of the query network extracted from it. We denote a sequence of query networks where each network in the sequence differs slightly from the preceding one as dynamically evolving queries. Existing network alignment methods consider each query network independently. Thus, the total cost of aligning all the query networks in a given sequence of dynamically evolving queries is equal to the sum of the alignment costs of all the individual networks in this sequence. This cost can be prohibitively expensive particularly for large queries as well as large sequence of queries. We develop an incremental network alignment method, named INQ, for aligning a sequence of dynamically evolving query networks to a target network. Unlike existing methods, INQ benefits from the alignment of each query network while aligning the next query network in the sequence. As a result, it cuts down the search space and thus the running time dramatically. To the best of our knowledge, this is the first study which aligns networks in an incremental fashion. Our extensive experiments on both real and synthetic networks demonstrate that INQ is orders of magnitude faster than existing methods such as ColT. We also demonstrate that the results of INQ are highly accurate even for very long sequence of query networks and INQ successfully finds functionally similar pathways in cross-species protein interactions networks.