Wine aroma results from the combination of numerous volatile compounds, some produced by yeast and others produced in the grapes and further metabolized by yeast. However, little is known about the consequences of the genetic variation of yeast on the production of these volatile metabolites, or on the metabolic pathways involved in the metabolism of grape compounds. As a tool to decipher how wine aroma develops, we analyzed, under two experimental conditions, the production of 44 compounds by a population of 30 segregants from a cross between a laboratory strain and an industrial strain genotyped at high density. We detected eight genomic regions explaining the diversity concerning 15 compounds, some produced de novo by yeast, such as nerolidol, ethyl esters and phenyl ethanol, and others derived from grape compounds such as citronellol, and cis-rose oxide. In three of these eight regions, we identified genes involved in the phenotype. Hemizygote comparison allowed the attribution of differences in the production of nerolidol and 2-phenyl ethanol to the PDR8 and ABZ1 genes, respectively. Deletion of a PLB2 gene confirmed its involvement in the production of ethyl esters. A comparison of allelic variants of PDR8 and ABZ1 in a set of available sequences revealed that both genes present a higher than expected number of non-synonymous mutations indicating possible balancing selection. This study illustrates the value of QTL analysis for the analysis of metabolic traits, and in particular the production of wine aromas. It also identifies the particular role of the PDR8 gene in the production of farnesyldiphosphate derivatives, of ABZ1 in the production of numerous compounds and of PLB2 in ethyl ester synthesis. This work also provides a basis for elucidating the metabolism of various grape compounds, such as citronellol and cis-rose oxide.