Water absorption in wheat flour is a crucial parameter for optimizing bread-making processes. The determinants of wheat flour water absorption were investigated through the analysis of 28 compositional and technological properties of 150 wheats grown in France. A multiple linear regression approach was used to predict the water absorption, selecting the best model through successive examination of Bayesian Information Criterion, Variance Inflation Factor and minimizing the total number of variables. A model with protein content, soluble starch, damaged starch and specific viscosity from water extractable arabinoxylans was identified as the best trade-off between the number of variables and the predictive performances among all possible models. Soluble Starch, varying between 1.11 and 6.21 g/100 g flour a new criterion measured alongside water-extractable arabinoxylans content, varying between 0.26 and 0.86 g/100 g flour, shows significant potential to predict water absorption compared to damaged starch.