Significance In the context of genome and metagenome sequencing, the assignment of function to sequences is a serious issue. In the case of enzymes of strong substrate specificity, such as those involved in the breakdown of polysaccharides (e.g., glycoside hydrolases and polysaccharide lyases), assignments become unreliable when sequence similarity to experimentally studied enzymes is low. To better explore the sequence-to-function relationships of these enzymes, we successfully applied a strategy based on a rational bioinformatic selection of enzyme targets, synthetic gene synthesis, and screening of recombinant proteins on a wide diversity of carbohydrate substrates. Seventy-nine of our 564 targets exhibited enzymatic activity, including three activities that have not been described previously, and 13 novel enzyme families could be defined.