Review and prepare for lab
Design reports are key to set expectations for how a round will go, and pick the plates to commit to the lab.
Funnel and output overview
The report will show how many sequences were down-selected from the total pool, with stats on how many of those were unique. You also get a downloadable table of all variants. Mutation-frequency tables and a mutational-load graph are broken down per template.
Plate-level likelihood-of-success
A curve of the number of sequences (x-axis) against the likelihood that that many meet all constraints (y-axis). It reads as "≈20 sequences are nearly certain to meet all constraints; ~50% chance that 45 do; effectively impossible that more than ~65 do." It doesn't name which sequences — just how many. This is the panel for managing expectations about the plate as a whole.
Predicted fold improvement of the top-N
For the primary objective, the probability (y-axis) of a given fold or absolute improvement (x-axis) for the top-N variants. It tells you what to expect from your best candidates — e.g. "the top five are nearly certain to give a 16–18% decrease in activity, and very unlikely to exceed 25%."
Constraint-success plot
The most useful panel for diagnosing a hard-to-meet constraint. Columns show, at left, predicted fold improvement for the primary objective (binned), and to the right, the likelihood of meeting each constraint. You can see at a glance which constraints are nearly certain to be met and which run the full gamut from "almost certainly met" to "almost certainly not."
Tip: For good confidence in a constraint, look for roughly 80%+ likelihood of meeting it. If a constraint shows a broad spread and other constraints or your fold improvement are suffering, soften that constraint to give the model room. Loosening a secondary constraint can substantially raise primary performance (one advanced example moved primary activity from 1.5× to 2.7×).
Mutational load vs predicted performance
Predicted fold improvement against mutation count, split by template. Useful for seeing whether a template's designs performs better at a given mutation load.
Inter-assay correlation
Predicted fold improvement on the objective against the likelihood of meeting a constraint. Positive correlations mean two aims can be optimized together; a negative slope reveals a trade-off the model is forced to make (pushing the objective lowers the chance of meeting the constraint).
Preparing the dataset for the lab
Export Once you're ready, export the variants in the format your lab uses for synthesis and screening. The generated variants are available as a downloadable table.
Add your biological controls: Add them to the plate before sending to the lab.
Tip: Keep column names, units, and ID conventions identical to previous rounds so the measured results upload cleanly and combine with your accumulated data.