Fig. 4
From: Difficulty in inferring microbial community structure based on co-occurrence network approaches
Effects of the community type on co-occurrence network performance (AUPR value) in the cases of Pearson’s correlation-based method (a), Spearman’s correlation-based method (b), MIC-based method (c), SparCC (d), REBACCA(e), CCLasso (f), Pearson’s partial correlation-based method (g), Spearman’s partial correlation-based method (h), and SPEIC-EASI (i). Vertical-axis labels correspond to the community types: random community (random; blue), mutualistic community (mutual; orange), competition–mutualism mixture community (mix; green), competitive community (compt; red), and predator–prey (parasitic) community (pp; purple). Network size n = 50 and average degree 〈k〉 = 2. Random network structure was considered. smax was set to 0.5. The number of samples was set to 300. Error bars indicate standard deviations