Your best Round 4 variant has 3× activity — and 40% lower expression. Your +15°C Tm hit doesn't tolerate process pH. The two best single mutants combined perform worse than either parent. And your assay drifted between batches, so half your data is noise.

Directed evolution optimizes one property per round. Rational design from a crystal structure misses the distant positions that drive real-world performance. Combining single winners ignores epistasis. None of these approaches survive the jump from bench assay to bioreactor.

Cradle was built for teams past the screening hit. Our job is to get you to a variant that holds up.