Fig. 5
From: Difficulty in inferring microbial community structure based on co-occurrence network approaches
Relationship between co-occurrence network performance (baseline-corrected AUPR value) and the ratio of predator–prey (parasitic) interactions (1 – pC). As representative examples, Pearson’s correlation-based method (a), Pearson’s partial correlation method (b), CCLasso (c), and SPIEC-EASI (d) are shown. For each method, the following cases are shown: network size n = 50 and average degree 〈k〉 = 2; n = 100 and 〈k〉 = 2; n = 50 and 〈k〉 = 8; and n = 100 and 〈k〉 = 8. Random interaction matrices and random network structure were considered. smax was set to 0.5. The number of samples was set to 300. rs and p indicate the Spearman’s rank correlation coefficient and the associated p-value. The raw values (i.e., the values before averaging) were used for calculating rs