Protein concentration and composition are key components of the end-use value for wheat (Triticum aestivum L.) grain. Although the qualitative composition of the grain is genetically determined, the quantitative composition is significantly modified by growing conditions, and there are important management x genotype x environment interactions. We recently reported a model of grain N accumulation and partitioning for wheat grain. The main assumptions made in this model are: (1) the accumulation of structural/metabolic proteins (albumins-globulins and amphiphils) is sink-driven and is a function of temperature; (2) the accumulation of storage proteins (gliadins and glutenins) is supply limited; (3) on the one hand the allocation of structural/metabolic proteins between albumin-globulin and amphiphilic protein fractions and on the other hand the allocation of storage protein between gliadin and glutenin fractions during grain growth is constant. A modified version of this grain model has been coupled with a revised version of the wheat simulation model Sirius99, allowing us to analyze the interactions between the vegetative sources and the reproductive sinks for N at the crop level. The main modifications to Sirius99 concerned the post-anthesis N uptake and remobilisation. After anthesis, the potential rate of crop N uptake was assumed to decrease linearly with accumulated thermal time, and the actual rate of N uptake was limited by the capacity of the stem to store accumulated N. During grain filling the daily rate of N transfer to grain was calculated daily according to the current crop N-status. The coupled model (SiriusQuality1) simulated dynamics of total grain N and of the different grain protein fractions reasonably well. At maturity, measured total grain N ranged from 0.56 to 1.32 mg N grain(-1), and the observed and simulated total grain N were well correlated (r(2) = 0.82, slope = 1.08) with a mean error of prediction of 0.11 mg N grain(-1). The simulated kinetics of crop N accumulation and stem N were closer to the observations with SiriusQuality1 than with Sirius99, in particular during the reproductive stage. At maturity, simulated and observed quantities of albumins-globulins were poorly correlated (r(2) = 0.02). Over the 18 experimental cases studied here, the quantity of storage proteins varied more than three-fold, and the observed and simulated quantities of gliadins and glutenins were well correlated (r(2) = 0.79 and 0.72, respectively). The simulations of total N and storage proteins accumulation provided by SiriusQuality1 confirmed that accumulation of grain N is overall source-rather than sink-regulated, at least under non-luxury N conditions. SiriusQuality1 provides a simple mechanistic framework that explains environmental effects on grain protein concentration and composition. The next step is to incorporate genetically related model parameters that will portray genotypic differences in protein concentration and composition.