Bioconversion of plant derived lignocellulosic materials into biofuels has drawn increasing attention as an attractive strategy for replacement of fossil fuels. Saccharomyces cerevisiae, also known as baker's yeast, is considered one of the most promising organisms for ethanol production from lignocellulosic feedstock. Unfortunately, pentose sugars, which make up to 30% of biomass hydrolysate, cannot be utilized by S. cerevisiae. Heterologous pentose utilization pathways from bacterial and fungal resources have been introduced into S. cerevisiae to enable the assimilation of pentose sugars. However, pentose utilization of recombinant S. cerevisiae strains are inefficient due to the low expression level and activity of heterologous genes, redox imbalance resulted from different preference of cofactor for oxidation and reduction reactions, and suboptimal metabolic flux through different catalytic steps. A lot of research has been carried out to improve the pentose utilization by S. cerevisiae targeting different aspects of these issues, but it is very challenging to come up with a single strategy which can solve all three problems at the same time. Here we report the combinatorial optimization of a highly efficient pentose utilization pathway in S. cerevisiae. Specifically, we chose the fungal three-step xylose utilization pathway as our target pathway, and performed optimization of the pathway through shuffling of pathway enzymes or promoters. For the enzyme-based optimization, 10 to 20 different enzyme homologues from various fungal species with different catalytic efficiency and cofactor preference have been cloned for each pathway enzyme, and were assembled into a library of xylose utilizing pathways using our recently developed DNA assembler method. For the promoter-based optimization, promoter mutants with varied strength were created by error-prone PCR and randomly assembled into the pathway to generate a library of xylose utilizing pathways. The resultant libraries exhibited a very high efficiency for correct assembly of the complete multi-gene pathway and good diversity of different homologues within the same catalytic step. Using the libraries generated, clones with better combination of enzyme homologues can be selected by faster cell growth. This method can also be applied to directly search for the pathway with a better fit in industrial yeast strains, which may have different metabolic flux patterns compared to the laboratory S. cerevisiae strains.