Approximation to the Pareto set of optimum primary/promiscuous activities.

<p>(A) Values of the primary and promiscuous activities for 29 variants randomly selected from a combinatorial library based on the 10 mutations derived from SCA analysis (see text and <a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1002558#pcbi-1002558-g004" target="_blank">Figure 4</a>). Values for the wild-type thioredoxin from <i>E. coli</i> and the background P34H variant are also included. Note that logarithms of activities are used here in and also in all the other panels of this figure. (B) Partial-least-squares (PLS) reconstruction of the data for the whole combinatorial library (small red data points). The experimental data of panel A (used as a basis for the reconstruction) are also shown, although, for the sake clarity we have omitted the errors bars here (as well as in panels C and D). The reconstructed data (red points) are actually derived from 20 bootstrapping replicas (see <a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1002558#s3" target="_blank">Methods</a> for details). (C) Prediction of the Pareto set from the PLS-reconstructed data. The 11 variants belonging to this predicted Pareto set are shown with blue circles. (D) The actual experimental activity values for the 11 variants are shown (open green squares). (E) Expanded experimental variant set including the original 29 variants (panel A), the wild-type thioredoxin from <i>E. coli</i>, the P34H background variant and the 11 variants added as a result of PLS-reconstruction/Pareto-set-prediction (panels C and D). (F) The actual Pareto set of the expanded variant set is shown (green data points).