Comparison of the MFPS and MDPS using the High Fidelity Ratio (HFR). HFRs at the A) OTU level and B) genus level are shown. Both MFPS A and MFPS B (OTU penalties set to 1, and genus penalties set to 0 and 30, respectively) score higher than the MDPS in all probe set sizes examined but the difference narrows from n = 20 to n = 80 probes. Randomly chosen probe sets perform less well. The Random 1000 probe sets selected probes from the top 1000 most conserved probes and the Random 4000 from the top 4000. The MFPS and MDPS also selected probes from the 1000 most conserved probes. Error bars are the standard deviations of 100 probe sets per data point.