The high cost of enzymes for biomass deconstruction is a major impediment to the economic conversion of lignocellulosic feedstocks to liquid transportation fuels such as ethanol. We have developed an integrated high throughput platform, called GENPLAT, for the discovery and development of novel enzymes and enzyme cocktails for the release of sugars from diverse pretreatment/biomass combinations. GENPLAT comprises four elements: individual pure enzymes, statistical design of experiments, robotic pipeting of biomass slurries and enzymes, and automated colorimeteric determination of released Glc and Xyl. Individual enzymes are produced by expression in Pichia pastoris or Trichoderma reesei, or by chromatographic purification from commercial cocktails or from extracts of novel microorganisms. Simplex lattice (fractional factorial) mixture models are designed using commercial Design of Experiment statistical software. Enzyme mixtures of high complexity are constructed using robotic pipeting into a 96-well format. The measurement of released Glc and Xyl is automated using enzyme-linked colorimetric assays. Optimized enzyme mixtures containing as many as 16 components have been tested on a variety of feedstock and pretreatment combinations. GENPLAT is adaptable to mixtures of pure enzymes, mixtures of commercial products (e.g., Accellerase 1000 and Novozyme 188), extracts of novel microbes, or combinations thereof. To make and test mixtures of ?10 pure enzymes requires less than 100 ?g of each protein and fewer than 100 total reactions, when operated at a final total loading of 15 mg protein/g glucan. We use enzymes from several sources. Enzymes can be purified from natural sources such as fungal cultures (e.g., Aspergillus niger, Cochliobolus carbonum, and Galerina marginata), or they can be made by expression of the encoding genes (obtained from the increasing number of microbial genome sequences) in hosts such as E. coli, Pichia pastoris, or a filamentous fungus such as T. reesei. Proteins can also be purified from commercial enzyme cocktails (e.g., Multifect Xylanase, Novozyme 188). An increasing number of pure enzymes, including glycosyl hydrolases, cell wall-active esterases, proteases, and lyases, are available from commercial sources, e.g., Megazyme, Inc. (www.megazyme.com), NZYTech (www.nzytech.com), and PROZOMIX (www.prozomix.com). Design-Expert software (Stat-Ease, Inc.) is used to create simplex-lattice designs and to analyze responses (in this case, Glc and Xyl release). Mixtures contain 4-20 components, which can vary in proportion between 0 and 100%. Assay points typically include the extreme vertices with a sufficient number of intervening points to generate a valid model. In the terminology of experimental design, most of our studies are "mixture" experiments, meaning that the sum of all components adds to a total fixed protein loading (expressed as mg/g glucan). The number of mixtures in the simplex-lattice depends on both the number of components in the mixture and the degree of polynomial (quadratic or cubic). For example, a 6-component experiment will entail 63 separate reactions with an augmented special cubic model, which can detect three-way interactions, whereas only 23 individual reactions are necessary with an augmented quadratic model. For mixtures containing more than eight components, a quadratic experimental design is more practical, and in our experience such models are usually statistically valid. All enzyme loadings are expressed as a percentage of the final total loading (which for our experiments is typically 15 mg protein/g glucan). For "core" enzymes, the lower percentage limit is set to 5%. This limit was derived from our experience in which yields of Glc and/or Xyl were very low if any core enzyme was present at 0%. Poor models result from too many samples showing very low Glc or Xyl yields. Setting a lower limit in turn determines an upper limit. That is, for a six-component experiment, if the lower limit for each single component is set to 5%, then the upper limit of each single component will be 75%. The lower limits of all other enzymes considered as "accessory" are set to 0%. "Core" and "accessory" are somewhat arbitrary designations and will differ depending on the substrate, but in our studies the core enzymes for release of Glc from corn stover comprise the following enzymes from T. reesei: CBH1 (also known as Cel7A), CBH2 (Cel6A), EG1(Cel7B), BG (?glucosidase), EX3 (endo-?1,4-xylanase, GH10), and BX (?-xylosidase).