Figure 2
From: The chemical signatures underlying host plant discrimination by aphids
Results of Random Forest analyses and correlation of abundance in top ranking m/z bins with LD1 profile. A-D; Rank variability (interquartile range) as a function of median rank for the RF importance score (mean decrease in Gini coefficient) of the top 100 m/z bins from 500 runs of the RF regression models: LD1 discrimination profile against (A) polar and (B) non-polar metabolic data and LD1 overall acceptance profile against (C) polar and (D) non-polar metabolic data. m/z bins with the highest average rank best explain variation in LD1 profile and greater confidence can be placed in bins with low rank variability. Points in black are key m/z bins used for further investigation. (E–L) Correlations between key m/z bin values highlighted by RF models and LD1 scores. Mean ± SEM, Significance tested using Spearman’s rank correlation with FDR correction: NS – not significant, *P < 0.05, **P < 0.001, ***P < 0.001.
