Current directed evolution strategies for building and screening libraries generally involve generating protein sequence diversity randomly across the whole sequence or in controlled random fashion at defined positions within the protein.  These libraries generally have a large number of members that are “negative” with respect to the primary property, and require large numbers (10⁴ to 10⁶) of members be screened in order to find the relatively small number of positive mutations.  As positive mutations accumulate for one property, all others get worse.  We have created a method to build libraries that optimize several properties at once and only require screening of 100’s of members to discover proteins with the desired properties.  This method uses a survey of sequence activity relationships for all properties as part of the design process and provides a rapid and efficient way to engineer proteins for multiple properties.