In a similar way to the binary code of ones and zeroes that tell a computer program what to do, living cells follow instructions encoded in DNA to construct organisms. It's long been thought that any given DNA sequence would always create the same protein every time a sequence is read out by the cell, but now researchers have found the first exception to this "universal rule," in a species of yeast that chooses between two different translations each time.

DNA molecules are made up of four chemical bases, represented by the letters A, C, G and T. The cells of living organisms consult this genetic instruction manual to build proteins – the workhorses of the body. The DNA code is read in groups of three bases collectively called codons, where each codon always creates a specific amino acid. GGA, for example, encodes for glycine. That means researchers can read DNA and accurately determine which amino acids – and by extension which proteins – are built from each sequence.

At least, that's how it's always seemed to work previously. But now, researchers at the Max Planck Institute and the University of Bath have identified a microbe that breaks the universal rule. That makes it impossible to predict exactly which protein will be created at any given time, no matter how closely you study the DNA code.

The trend-bucking organism belongs to a group of yeast species that are already a bit strange. In humans and most other living things, the codon CTG translates to the amino acid leucine. But in this yeast family, some members use the sequence to create serine, while others make alanine out of it.

So far, so good. These kinds of differences between species are rare, but not unheard of. Where things get really weird is in the black sheep of the family – a yeast species known as Ascoidea asiatica. From the one codon, this organism will encode one of two amino acids, apparently at random.

"This is the first time we've seen this in any species," says Laurence Hurst, co-author of the study. "We were surprised to find that about 50 percent of the time that CTG is translated as serine, the remainder of the time it is leucine. The last rule of genetics codes, that translation is deterministic, has been broken. This makes this genome unique – you cannot work out the proteins if you know the DNA."

The researchers traced the coin-toss mechanism to the tRNA translator molecules, which also seem to be strangely binary in A. asiatica. But the real question is what benefit the yeast gets out of this randomization, and the short answer seems to be – none, really.

"Swapping a serine for leucine could cause serious problems in a protein as they have quite different properties – serine is often found on the surface of the protein whereas leucine is hydrophobic and often buried inside the protein," says Stefanie Mühlhausen, co-author of the study. "We looked at how this strange yeast copes with this randomness and found that A. asiatica has evolved to use the CTG codon very rarely and especially avoids key parts of proteins."

In the end, it seems the universal rule is universal for a reason. Using a molecular clock the researchers estimated that the random encoding trait could be around 100 million years old, but related species seem to have evolved a more predictable pattern to be safe. The uniqueness of A. asiatica might therefore represent an evolutionary dead end.

The research was published in the journal Current Biology, and the team describes their work in the video below.