Abstract: It has long been known that many proteins require folding via molecular chaperones for their function. Although it has become apparent that folding imposes constraints on protein sequence evolution, the effects exerted by different chaperone classes are so far unknown. We have analyzed data of protein interaction with the chaperones in Saccharomyces cerevisiae using network methods. The results reveal a distinct community structure within the network that was hitherto undetectable with standard statistical tools. Sixty-four yeast chaperones comprise ten distinct modules that are defined by interaction specificity for their 2,691 interacting proteins. The classes of interacting proteins that are in turn defined by their dedicated chaperone modules are distinguished by various physiochemical protein properties and are characterized by significantly different protein expression levels, codon usage, and amino acid substitution rates. Correlations between substitution rate, codon bias, and gene expression level that have long been known for yeast are apparent at the level of the chaperone-defined modules. This indicates that correlated expression, conservation, and codon bias levels for yeast genes are attributable to previously unrecognized effects of protein folding. Proteome-wide categories of chaperone-substrate specificity uncover novel hubs of functional constraint in protein evolution that are conserved across 20 fungal genomes.