I am sure that Arcticio will never see a mammoth. They might see an elephant with mammoth DNA but a mammoth will never come back. That is the big problem about the DNA codes for an organism viewpoint – it is the same as a blueprint codes for my house. It doesn’t unless I have a builder. So unless someone finds a mammoth egg that is not going to happen.

As for stability – it depends on storage. A living organism is very environment dependent. A nutrient free organism is a dead one but my ssd does not need a sandwich or a drink. The shuttles used regular hard drives for storage and they took more than a few shocks and could be restored, organic storage is much less effective and also much more error prone at making subjective interpretation of observations and data.

I am wondering how you can fiind out the DNA replication error rates at a level which is below the accuracy of the measurement device (the sequencing device itself). Surely that is impossible. I cannot measure to the nearest nanometre on a metre rule. Somatic mutation rates are difficult to estimate and we always sequence populations. There is a move in recent papers to sequence single cells to compare them for circulating cancers for example and there can be large differences.

-Wang-Lo.

If the data is stored in e. coli in some place in the genome that would not affect their viability, the entire 6.3 megabit would not fit in the 4.2Mbp e. coli genome. But the book could be spread over as many individual bacteria as necessary. According to the blog post at http://sandwalk.blogspot.co.nz/2007/07/mutation-rates.html the rate of neutral mutation (mutations that have no positive or negative effect on viability) in e. coli after accounting for the DNA repair mechanism is approximately 1 out of every 1e10 nucleotide base pair replications. Assuming 24 hours per generation (a more realistic real world number than the faster replication under laboratory conditions) and approximating 365.25 days in a year, that is about 3.6 million generations, or a total of 6.3 million times 3.6 million nucleotide replications, for an estimated number of errors at the end of 3.6e6 times 6.3e6 times 1e-10 which is 2268 errors in each 6.3 million bit copy of the book. or less than a 0.04% error rate. That is a tiny amount that can be compensated for with a simple error correction code that would not add many bits to the total. By comparison, the Reed-Solomon code used for the first stage error correction of an audio CD requires 32 bits in all to encode 28 bits of data, and it can correct up to 2 bytes of error out of every 32 byte block.

So if you could somehow arrange to maintain the proper environment to support the colony of e. coli for 10,000 years and have someone around to decode them at the end, the book should be readable.

[CZ: Just to clarify--Church and his colleagues did not insert the DNA into a living organism. The molecule sits in a chip.]