Fig. 4

Flow diagram of the classification of proteins of the GOS dataset into families and representatives. First, all protein sequences were annotated using HMM-profiles obtained from PFAM and TIGRFAMs. Proteins that did not match HMMs with scores below the selected threshold were clustered de novo using MCL. Resulting MCL-based classes of small size were excluded. For each HMM-based and MCL-based family that contained sufficient complete sequences, a representative was defined. In this way 9,771 representatives standing in for 4,969,723 proteins were assigned. This set of representatives can then be used to create a custom expression library, which can be screened for a desired target activity