The last paragraph of that article proves the OP's point, you can't do it by working from first principles.

"Even Venter acknowledges that syn3.0’s genome, although new, was designed by trial and error, rather than being based on a fundamental understanding of how to build a functioning genome."