Slashdot Mirror
Genetic Database Hits One Billion Entries
ChocSnorfler writes to tell us that the Sanger Institute is reporting that their Genetic Record Database has hit one billion entries, making it the world's largest. From the announcement: "The Trace Archive is a store of all the sequence data produced and published by the world scientific community, including the Sanger Institute's own prodigious output as a world-leading genomics institution. To grasp how much data is in the Archive, if it were printed out as a single line of text, it would stretch around the world more than 250 times. Printing it out on pages of A4 would produce a stack of paper two-and-a-half times as high as Mount Everest. The Archive is 22 Terabytes in size and doubling every ten months."
Some dumbass is always printing 300 pages of documents and hogging the printer. Forchrissakes, just figure out what pages you need and print those! Asshole.
The amount of data here is really enormous. To put it in perspective, if you lined up 7143 blondes, the number of strands of hair present would approximately equal the number of entries in this database.
"To grasp how much data is in the Archive, if it were printed out as a single line of text, it would stretch around the world more than 250 times. Printing it out on pages of A4 would produce a stack of paper two-and-a-half times as high as Mount Everest. "
I have twice that much data on my 128k thumbdrive, if printed out in 72 point font size.
Anyone care to translate this into volkswagens, or libraries of congress?
Wow, that's almost 12U of rack space. Oh my *yawn*
Now the fact that that's all genetic data, that's amazing considering a human is only ~1GB so 22,000 humans worth.
- Adam L. Beberg - The Cosm Project - http://www.mithral.com/
if it were printed out as a single line of text, it would stretch around the world more than 250 times. Printing it out on pages of A4 would produce a stack of paper two-and-a-half times as high as Mount Everest.
Such claims should be taken with a grain of salt until they reveal what fonts and point sizes they use.
if it were printed out as a single line of text, it would stretch around the world more than 250 times. Printing it out on pages of A4 would produce a stack of paper two-and-a-half times as high as Mount Everest
Did anybody else think "Wow, I've got a great idea for a mural for the space elevator!"
Anybody?
Uh, well, it's late...
--MarkusQ
Would somebody please torrent it?
If we stacked up all of the useless length metaphors/comparisons from end to end, they'd still add up to a non-useful mental image of a billion genetic records.
I mean, "printed out as a single line of text, it would stretch around the world more than 250 times" means what, in terms of helping us picture this? I take it that we're not supposed to be able to imagine a billion records, but we can all clearly picture some text wrapped around the planet 250 times? Ah, that's much more helpful!
Now, I just got done re-indexing 10 million records in a database, and I can sort of picture 100 times that much work. This is slashdot! More nerdly examples, please.
Don't disappoint your bird dog. Go to the range.
When we figure out what all of that does. For every organism as or more complex than your average bacterium, there's a large amount of what amounts to filler DNA. Viruses don't have this problem, as few of them are large enough to even get by without overlapping reading frames. If you shrink this dataset down to only sequences that encode functional proteins (read: genes), there's still an insane amount of information. If you then remove the introns, the dataset gets even smaller. But of course, we don't really know if the introns and intra-genic regions of DNA (the so-called 'junk DNA') have functions (or how many they have), although some do act as regulators of transcription.
Given that a change of just 1 base in 500 of the 16S rRNA gene is sufficient to differentiate between two different species of bacteria, I have to wonder how many of these entries are quasi-redundant. When you consider how many species of bacteria are known to man, that means that there are literally thousands of potential entries for each gene. Unless, of course, they're storing only consensus sequences, which still vary widely between genera.
Sadly, the trend here seems to be more of 'sequence it, upload it, and patent it' instead of 'sequence it, upload it, figure out what it does/makes, do something useful with it'. Knowing the sequence for the Ubiquitin gene is all well and good, but it's of little practical importance. Being able to construct designer proteins to treat illnesses based on that information, however, is a truly worthy goal. Unfortunately, that's also where the 'patent it' part comes into play...
From: http://www.learner.org/channel/courses/biology/tex tbook/genom/genom_7.html
A biological approach to determining the function of a gene is to create a mutation and then observe the effect of the mutation on the organism. This is called a knockout study. While it is not ethical to create knockout mutants in humans, many such mutants are already known, especially those that cause disease. One advantage of having a genome sequence is that it greatly facilitates the identification of genes in which mutations lead to a particular disease.
The mouse, where one can make and characterize knockout mutants, is an excellent model system for studying genetic diseases of humans; its genome is remarkably similar to a human's. Nearly all human genes have homologs in mice, and large regions of the chromosomes are very well conserved between the two species. In fact, human chromosomes can be (figuratively) cut into about 150 pieces, mixed and matched, and then reassembled into the 21 chromosomes of a mouse. Thus, it is possible to create mutants in mice to determine the probable function of the same genes in humans. Genetic stocks of mutant mice have been developed and maintained since the 1940s.
One goal of the mouse genome project is to make and characterize mutations in order to determine the function of every mouse gene. After a particular gene mutation has been linked to a particular disorder, the normal function of the gene may be determined. An example of this approach is the mutated gene that resulted in cleft palates in mice. The researchers found that the gene's normal function is to close the embryo's palate. An understanding of the genetics behind cleft palate in mice may one day be used to help prevent this common birth defect in humans.
In most cases they work backwards. You start with a known protein, determine its amino acid sequence, and then convert that into the most likely DNA sequence (accounting for codon bias). Primers/probes are then generated for the 3' and 5' ends of the probable DNA sequence. If you're working with a small genome like that of a bacterium, you can perform a restriction digest to get random hunks of chromosome. These are then amplified via PCR using your designer primers. The final product is then sequenced.
In other cases you can create a gene knockout by splicing a random gene into your gene of interest. This causes your target gene to encode a non-functional protein. Then you watch and see what happens to the test subject. In some cases the creature dies because the gene turned out to be extremely important. In others it results in minor to significant impairment. But because of the complexity of most organisms, single-gene knockouts usually don't have too much effect - the creature has multiple pathways that can accomplish the same goal. This is especially true for critical functions like those in the immune system.
Printing it out on pages of A4 would produce a stack of paper two-and-a-half times as high as Mount Everest.
You can't do that with ordinary A4 paper. You need to reinforce it on the sides at least so it won't tumble over. Plus, I doubt the paper would sit still with the high winds once it gets above a few thousand feet. Sheesh.
I mean, most of that data is just redundant pairs of A-G C-T T-G etc...
I reckon you could zip it up and it'll fit on a couple of floppy disks.
Anyway, the point is you are not about to be able to fit a genome on a floppy disk. Not even close.
Most exciting phrase in science: not "Eureka!" but "Hmm... That's funny..." -Asimov (abridged for \. limits)
I'm not impressed. I already have genetic material all over my computer.
(Oops, did I just admit something bad?)
All this hype about how vastly much paper you get if you print it all out misses the wonder of the thing.
The wonder isn't how BIG the human genome is - the amazing thing is how *TINY* it is.
The human genome is 3 billion base pairs...each base pair is one of only four possibilities - so two bits each. 750 Megabytes...that's one CD-ROM. There is a lot of redundancy in it too - many of those base pairs are never 'expressed' as proteins, many are replicated redundantly dozens of times. So with compression, or even just deleting the junk - you'd get it down to maybe 100 to 200 megs - tops.
I find it utterly amazing that all that complexity is so amazingly compactly encoded.
Yeah - that's a lot of bits of paper - or 600 floppy disks or some other bullshit - but by the standards of modern media, it's MICROSCOPIC.
Announcements like this would do better to explain how LITTLE data this really is - that's the wonder of the thing.
www.sjbaker.org
...the entire database would fit on just one sheet of A(-24) paper. (Yes, I actually did the math.)
There's no failure quite as dissatisfying as a complete and total solution to the wrong problem.
I won't give away the ending, but my favorite part is:
ctattggacttggaatcggatattggacacttggaatcggata
Go FoxPro!
All your base (pairs) belong to us.
The signal data is composed of peaks and troughs across 4 channels, corresponding to the 4 base types. A peak in a channel corresponds to a base of that type passing in front of the detector. In your typical sampling configuration, a peak is made up of about 12 data pts.
Now, since each sampled point in the signal is stored as a 4 byte int and the base for that peak is stored as a 1 byte char, then you've got basically a 192:1 ratio of techincally superfluous signal data to actual DNA sequence.
Since there are yet other peices of information in the file, this ratio is actually larger.
Of course, there is a good reason for keeping trace data rather than just the DNA sequences, the notion being that you have more information with which to validate the integrity of what you've done. There have been cases where scientific databases have had their data integrity damaged over time by low quality (ie. mistakes) submissions.
In this case, they're retain the wrong file type, as it doesn't store the original unfiltered data signal, only a heavily filtered and manipulated one. Most modern basecallers start from the original unfiltered data to gain more advantage through better processing, you cannot do this with the file type they are retaining.