Fig. 3

a The distributions of Q3, SOV’99, and SOV_refine from a sample of 14,252,776 scores calculated from Top8000 database. b The probabilities (P-values) of having a given score of Q3, SOV’99, and SOV_refine greater than or equal to a certain value. We can find that the probabilities of finding Q3 ≤ 0.26, SOV’99 ≤ 0.24, and SOV_refine ≤ 0.19 for two random proteins are close to 1. These findings indicate that we can distinguish two protein structures or models at the secondary structure level if their Q3 or SOV scores are greater than or equal to these corresponding thresholds