This is not how they conduct gain of function research. It's uses serial passage (using biological hosts and the virus's own variations). This produces vary little evidence of manipulation.