Construction of synthetic genes is today the most elegant way to optimize the heterologous expression of a recombinant protein. However, the selection of positive clones that incorporate the correct synthetic DNA fragments is a bottleneck as current methods of gene synthesis introduce 3.5 nucleotide deletions per kb. Furthermore, even when all predictable optimizations for protein production have been introduced into the synthetic gene, production of the protein is often disappointing: protein is produced in too low amounts or end up in inclusion bodies. We propose a strategy to overcome these two problems simultaneously by cloning the synthetic gene upstream of a reporter gene. This permits the selection of clones devoid of frame-shift mutations. In addition, beside nucleotide deletion, an average of three non-neutral mutations per kb are introduced during gene synthesis. Using a reporter protein downstream of the synthetic gene, allows the selection of clones with random mutations improving the expression or the folding of the protein of interest. The problem of errors found in synthetic genes is then turned into an advantage since it provides polymorphism useful for molecular evolution. The use of synthetic genes appears as an alternative to the error-prone PCR strategy to generate the variations necessary in protein engineering experiments.