How would you compare candidate protein variants?

Why

Define the application before ranking: a therapeutic binder, a screening reporter and a catalyst require different endpoints. Produce candidates under comparable conditions, record total and soluble expression, and determine whether functional assay inputs contain equivalent active protein. If possible, use concentration-response curves rather than a single endpoint and check assay linearity.

For enzymes, distinguish kcat, Km and catalytic efficiency under the relevant temperature, pH and substrate concentration. For binders, examine affinity, specificity, aggregation and developability in addition to apparent signal. For example, variant A can produce twice the raw reaction signal only because it yields twice as much soluble enzyme.

Its specific activity may be unchanged. Include positive reference variants, negative controls and interday repeats. Use uncertainty and application-weighted constraints to nominate a Pareto set instead of declaring one model score the global optimum.