The digital and analog worlds of protein engineering

Nature Biotech Correspondence Link

5 years ago if you seriously wanted a good protein, you wouldn't start with computational modeling, but in the last year lines became more and more blurred. Today, if i want a better protein X, do i start with a model or library design?  How would each of them explore the sequence space? How would it bias the search? What would be faster? What types of tasks should prioritize one over the other? We don't know! This transition period requires a bunch of ad hoc integrations, especially in industry.