They’ve reconstructed the genomes for the woolly mammoth and the Neanderthal — now paleogeneticists led by Ian Holmes at UC Berkeley have sequenced the earliest mammal genome. How do you do that without a primitive mammal handy? It’s a statistics hack; they extrapolate from the genomes of living animals. Writes Holmes,
The cool thing is that you can get a lot of information about ancestral genomes just by crunching probabilities — even if you don’t have any fossils, or mosquitos-trapped-in-amber, or time machines, or whatever.
The even cooler thing (to my bioinformatics-geek mind) is that the algorithms used for this are almost exactly the same ones that linguists use to reconstruct ancient languages, like “Indo-European” or “Gondwanese”.
The problem here is to construct a phylogenetic tree (an evolutionary family tree), then deduce the amino acid sequences of the common ancestral proteins at the branches. So by looking at the DNA of present-day mammals, we can infer the shared genome of a common ancestor. This is reminiscent of the divergence of words from their original Indo-European ancestors: for example, the Proto-Indo-European word for snow (*sneigwh-) can be reconstructed from the descendant words for snow in the descendant Indoeuropean languages (German schnee, French neige, Irish sneachta, Russian sneg, Sanskrit snihyati, and so on).
The first scientists who proposed to reconstruct the genetic past from the present were Linus Pauling and Emile Zuckerkandl, who, in 1963, resurrected an ancient form of hemoglobin. They actually reconstructed the molecule in the lab. Which leads us to wonder, now that we have entire genomes of ancient organisms, whether some kind of Crightonesque resurrections might be possible. Would you want to meet a Neanderthal?
Much more here, in a book on ancestral sequence reconstruction.