The aim of the study was establish if methane (CH4) yields (g/kg DM intake, DMI) could be predicted for individual cattle from rumen volatile fatty acid (VFA) molar proportions, and animal (DMI) and feed (forage:concentrate ratios and ME content) variables. The dataset available consisted of 284 measurements (from 6 experiments) for beef cattle of CH4 yields measured in indirect respiration chambers. A compositional modelling approach was employed where VFA measurements were considered as a whole, instead of in isolation, emphasizing their multivariate relative scale by means of a log-ratio coordinate representation. The analysis revealed expected close groupings of acetate and butyrate; propionate and valerate; iso-butyrate and iso-valerate. Linear mixed models were then fitted to investigate relationships between CH4 yield and VFA (represented by meaningful log-contrasts of its components called balances) while accounting for other animals and diet-related covariates. A model including a normalized balance representing (acetate + butyrate) / propionate best explained the contribution of VFA to CH4 yield variation. DMI, forage:concentrate proportion (expressed as a categorical variable diet-type: high concentrate, mixed forage:concentrate or high forage) and diet ME concentration were also statistically significant covariates. Our results provided new insights into the inter-relationships amongst VFA measurements and also between VFA and CH4 yield. In conclusion, it is possible to use VFA molar proportions to predict CH4 yields from individual cattle.