Design Characteristics:
Number of Varying Positions: 0Protein Characteristics:
Deleterious Mutations: 0%Assay Characteristics:
Assay noise: %Percent of clones with at least 1 deleterious mutation. These clones are a load for screening in any throughput.
Correlation between "Real" and Measured/Transformed activities are indicated here
Precision and Recall are standard measures used to asses models
Precision is defined as:
This is calculated based on retrieving top 10 truly active clones from the assay. Measured data is always transformed from true data through noise, surrogate correlation etc..
This is calculated based on retrieving top 10 truly active clones from predicted activity from modeling. These account for variations in measurements and surrogates through analyzing the data from an additive model. Works best when the variants are created with systematic representation of the substitutions. ProteinGPSTM from ATUM synthesizes this set as infologs to best extract information for Engineering the true activity
The ability to select top substitutions from the set is reflected here. Note how the "truly best" substitutions stays on top right, even when there is a lot of noise in the assay.
Note: Assay Noise can be reduced by replicates. This requires measuring a lot of samples in replicates. In turn, increasing the costs. Costs can be controlled switching to a surrogate assay, which in turn results in poor correlation with real assay.