Fig. 3: Variable selection for the MGR prediction model. | Nature Microbiology

Fig. 3: Variable selection for the MGR prediction model.

From: Soil microbiome indicators can predict crop growth response to large-scale inoculation with arbuscular mycorrhizal fungi

Fig. 3

a, Soil parameters were initially filtered through the removal of co-correlated variables, followed by random forest, stepAIC and glmulti analyses resulting in a final set of 15 variables, displayed as loading vectors in the PCA plot of all soil parameters. b, Soil fungal OTUs were selected using indicator species, differential abundance and random forest analyses, and a further refinement step using glmulti. This resulted in a final set of 13 sOTUs, depicted as loading vectors in the partial (adjusted for year effect) dbRDA plot, showing a clear grouping by MGR category. c, Establishment success of the inoculated AMF. Fields are shown in descending order of MGR (grey bars, representing the confidence interval of MGR for each field, displayed on the secondary y axis). AMF establishment success (displayed on the primary y axis) is shown as the difference (Δ) between inoculated and control samples for the relative abundances of SAF22 rOTUs and total colonization. The plot shows that there is no relationship between MGR and establishment success (indicated by the smoothed lines, which follow a different trend from that of MGR). However, the relative abundance of SAF22 and total root colonization are strongly correlated (see Extended Data Fig. 5c for pairwise correlations).

Back to article page