Unlocking the secrets of the human genome would be impossible without the computerized manipulation of massive amounts of data, including the majority of the three billion chemical units that comprise our own species' genetic blueprint. But what this "bioinformatics" revolution has provided, above all, is stark confirmation of the evolutionary basis of all life on Earth.
Sequence data, whether from proteins or nucleic acids, are well suited to computer processing because they are easily digitized and broken down into their constituent units. Simple computer programs can compare two or more strings of these units and evaluate degrees of similarity, search huge databases to match new sequences against known ones, and cluster groups of sequences in the form of a family tree.
The implications of research on the first proteins to be studied almost half a century ago were profound. These sequences were all rather small--insulin has only about 50 amino acids, depending on the species--but the variation between species was clear.
My own interest began with one of these simple molecules 40 years ago, when I was a postdoctoral student in Sweden. Fibrinopeptides are short sequences that are relatively easy to purify and have the virtue of changing significantly from species to species. As a result, we were able to show a strong correspondence between the fossil record and most of the changes that were observed in the fibrinopeptide sequences. So it was obviously possible to interpret the evolutionary past in terms of existing genetic sequences.
But advances in computing were indispensable to further progress. In 1965, Robert Ledley began the first real sequence database, the Atlas of Protein Sequence and Structure. In 1967, researchers produced a genetic tree of a score of animals and fungi that had virtually the same branching order as would have been drawn by a classical biologist, even though their computer was utterly ignorant of the comparative anatomy, paleontology, embryology, and other non-molecular attributes of these creatures. Finally, in 1970 a splendid computer innovation enabled the proper alignment of amino acid sequences (which is vital to all subsequent data management).
The interpretation of sequencing data then developed along two dimensions. First, there was a natural interest in the relationships between organisms. The assumption was that random changes occur along all limbs of a genetic tree, but depending on the protein, only some small fraction survives. If these survival rates were constant, then distances separating existing sequences could be calculated. A second kind of comparison focused on so-called paralogous proteins, which are descended from a common ancestor within the same creature as a result of gene duplications.
Both types of comparison showed that new proteins come from old ones, just as evolutionary theory would predict. Duplications of parts of a DNA genome occur constantly in all organisms, mainly as a result of random breakage and reunion events. Most of these duplicated segments are doomed to oblivion, because any proteins their genes produce are redundant. Occasionally, however, a slightly modified gene product proves adaptively advantageous, and a new protein is born. Often its function is very similar to the old one, but occasionally a drastic change occurs.
Then, in 1978, DNA sequencing came into wide use. Almost immediately, a flood of fresh genetic information overwhelmed the existing protein sequence database. A second storehouse, GenBank, was established, but initially it concentrated exclusively on DNA sequences. And yet the interesting information resided in the translated DNA sequences, that is, their protein equivalents.
It was one of those rare moments of opportunity when an amateur could compete with professionals. So I began my own database, mostly using translated DNA sequences; I called it NEWAT (New Atlas). Armed with a very primitive computer and some very simple programs written by an undergraduate student, we began matching every new sequence against all previously reported sequences and found many wholly unexpected relationships. By the time the Human Genome Initiative was launched at the end of the 1980's, the amount of data was no longer the limiting factor in the development of new knowledge; suddenly, managing it was.
Many scientists were skeptical about the human genome project. The human genome, they pointed out, contained a hundred times more amino acid sequences than the existing databases. So how would the genes be identified? How can you match up something that's never been found?
But every gene in a genome is not an entirely new construct, and not all protein sequences are possible--otherwise, the number of different sequences would be vastly greater than the number of atoms in the Universe. Only a miniscule fraction of possible sequences has ever occurred, through duplication, multiplication, and modification of a small starter set of genes. As a result, most genes are related to other genes.
I was confident that bioinformatics would enable us to identify all genes merely by sequence inspection. But after the completion of the first dozen microbial genomes, about half the genes remained unidentified--a level that has persisted through the first hundred genomes to be completed, including the human genome. Even one of the most studied organisms, E. coli , has an abundance of genes whose function has never been found.
Still, the benefits of deciphering genomes have been tremendous. The promises of quick medical applications may have been over-stated. But the inherent value is immeasurable: the ability to grasp who we are, where we came from, and what genes we humans have in common with the rest of the living world.