Figure 3

Similarity search throughput time for different minimal Tanimoto similarity. Explanation of graph: one hundred randomly sampled compounds were used to query for all similar compounds, repeated for different minimum Tanimoto similarities. Searches were done in the PubChem sample of 3.5 million compounds. As the similarity cutoff increases, performance goes up: finding all compounds similar to compound X with at least a Tanimoto similarity of 0.8 (= 80% similar) is faster than finding (many more) compounds that are 50% similar.