Report on the first contest on graph matching algorithms for pattern search in biological databases

Abstract

Graphs are a powerful data structure that can be applied to several problems in bioinformatics, and efficient graph matching is often a tool required for several applications that try to extract useful information from large databases of graphs. While graph matching is in general a NP-complete problem, several algorithms exist that can be fast enough on practical graphs. However, there is no single algorithm that is able to outperform the others on every kind of graphs, and so it is of paramount importance to assess the algorithms on graphs coming from the actual problem domain. To this aim, we have organized the first edition of the Contest on Graph Matching Algorithms for Pattern Search in Biological Databases, hosted by the ICPR2014 Conference, so as to provide an opportunity for comparing state-of-the-art matching algorithms on a new graph database built using several kinds of real-world graphs found in bioinformatics applications. The participating algorithms were evaluated with respect to both their computation time and their memory usage. This paper will describe the contest task and databases, will provide a brief outline of the participating algorithms, and will present the results of the contest.