Fig. 1

Variance components (A and B) and prediction accuracies (C and D) for methane emissions (A and C) and methane ratio (B and D) using different MRM. Variance components and prediction accuracies were estimated using models capturing the metagenomic relationship between samples calculated in four different ways: using Reference Based (RB) or Reference Free (RF) profiles that were each adjusted by taking the log (base 10) of proportions followed by normalization (L10) or by taking the log (base 10) of proportions followed by normalization within cohort (CA). The traits methane emissions (CH4) or methane ratio (CH4Ratio) were analyzed in three datasets: lambs fed lucerne pellets (lucerne lamb), lambs grazing pasture (grass lamb) and adults grazing pasture (grass adult). Prediction accuracy was estimated using cross-fold validation with each cohort a different fold