Figure 6

Parallel substructure search throughput time on 3.5 million compounds. Explanation of graph: sixty sampled substructures were used as a query for a parallel substructure search in the PubChem sample of 3.5 million compounds. The same searches were repeated with different maximum number of rows requested: 1000 (bottom), 10,000 (middle) and 100,000 (top). The graph displays the total throughput time in seconds but users can view the intermediate output generally faster as the search function spools ('pipes') results as they become available. Fastest throughput times are observed for query structures that commonly exist in compounds, resulting in high a success ratio of the VF2 isomorphism algorithm.