The article says they tried the additive/synthesis approach first, which didn't work. And then they tried the subtractive approach, starting with M. Mycoides. And in that case they were left with a lot of "unexplained" genes.
But he reminds us that the subtractive process entirely depends on the starting cell.
I would like to hear more about the difference between additive and subtractive methods -- it wasn't entirely clear from the article.
Additive method: Synthesize genomes with the genes you expect to be necessary, pretty much an optimization problem on the number of genes where you continually try to remove genes that were in previous iterations.
Subtraction method: After reading a bit of the paper, they utilized Tn5, which causes a DNA sequence to be arbitrarily inserted into the DNA (Random locations). This randomly disrupts genes, causing them to likely not function correctly.
Here's the logic:
If all genes were necessary we would expect no cells to live that had mutations.
If no genes were necessary we would expect all cells to have mutations at about an equal rate relative to the space they occupy. (Assuming no bias by the Tn5, but that's a nuance)
What they instead found was that some cells grew, but there were certain genes that were not mutated, meaning that they are likely necessary.
They also classified a few things like studying the growth of the cell (Slow growing, but still viable was classified as "quasi-essential") and implanting their minimal genome in another species, which failed giving evidence that a subset of genes that are sufficient for survival are not necessarily sufficient in all cells.
As far as I can tell, they synthesized the molecule.
A line from the paper...:
We used whole-genome design and complete chemical synthesis to minimize
I did fret for a second over what meaning of "remove"
daemonk was using (from the design? From the instantiation of the genome?) but decided I at least wasn't posting misinformation.
This reads to me as: They took a working system and did parts away until the system failed (or did stgh "odd" towards failure). Then they were looking at the pieces they took away and guessed what functionality these pieces have.
Shouldn't they have replaced them with a gene of the same length, filled with only stop codons?
Otherwise, it would be like cutting out instructions from an assembly language file, without knowing whether one of the instructions does something like adding a constant integer value to the program counter. You might think the code that was removed was critical to the operation of the program, but you could have instead replaced all the instructions with an equivalent number of no-ops, and it would still work as expected.
