From: Jennifer Mansfield-Jones <cardtris@umich.edu>
Date: Tue, 14 Jun 94 06:31:15 PDT
To: cypherpunks <cypherpunks@toad.com>
Subject: Re: DNA
In-Reply-To: <Pine.3.85.9406131137.A7527-0100000@cor.sos.sll.se>
Message-ID: <Pine.3.89.9406140939.A4507-0100000@pliny.ccs.itd.umich.edu>
MIME-Version: 1.0
Content-Type: text/plain


    For those who only look at the first screenful, a place to
go for fairly current details on gene sequencing is:

  Hillis, David M. and Moritz, Craig, eds.  1990.
     _Molecular Systematics_  Sinauer: Sunderland, MA.

    The most convenient way of keeping DNA is dried.  That, as I
understand it, is what the military are trying to do.  The idea
isn't, yet at least, actually to sequence it.  You don't need a
sequence for unambiguous identification.  The gimmick is RFLP:
restriction fragment length polymorphism.  You take a DNA sample
(in solution) from the unknown: say skeletal remains that might
be those of some MIA.  You expose that to enzymes that cut DNA in
specific locations depending on the DNA base-pair sequence of the
strands.  These enzymes are called restriction endonucleases --
hence the name of the technique.  Depending entirely on the DNA
sequence, the sample will get cut in a bunch of places giving a
bunch of DNA scraps of various different lengths.  You can get
chunks of different sizes to separate out by speed of movement
through a gel under an electric field.  According to preference,
you can then use either a stain or radioactive markers to tell
where in the gel the DNA fragments are.  If the pattern of
fragment migration is the same between the known and unknown, you
can now fit a name to the bones.  But, if the patterns aren't the
same, the DNA sequences the restriction enzymes looked for
weren't in the same places in the two samples.  That means they
couldn't have come from the same person.

    This is a bit of an oversimplification.  A lot of human DNA
has its restriction sites in the same places you'd find in apes,
never mind other humans.  Total DNA similarity between humans
and chimps is better than 90% overall.  Specific zones, called
hypervariable sequences, are the only ones really useful for
individual ID by DNA.  It also works very well for parentage
analysis.  So you might be able to identify an unknown sample
without a previous reference from that person if you could still
get samples from that individual's parents.

On Mon, 13 Jun 1994, Mats Bergstrom wrote:
> countries. These samples are usually frozen and saved for decades (for
> the purpose of comparison if the individual should fall ill; and for
> research if something might get interesting) at most laboratries. DNA-
> analysis efter thawing is no big deal with modern techniques. So if one

    The point I got a chuckle out of was the notion of freezing
blood samples as a routine thing.  To get much use at a molecular
level (either DNA or protein structure) out of frozen samples
over the long term (more than weeks) you have to keep it at -70C
or better.  People who study DNA are utterly paranoid about
freezer failure.  If they leave town, they may leave the cat
with an automatic feeder but they need someone to visit the
freezer once or twice a day and make sure it's okay.  If building
power fails (not that uncommon in old university science
buildings) you need a generator or a quick load of liquid
nitrogen to keep your frozen treasure from being ruined.  If
drying works, that's what will be used.

    I don't know, not being in that specialty myself, how good
the preservation quality of dry-stored DNA really is.  I can
easily imagine it being good enough for actual sequencing if it
had been quickly freeze-dried and stored under nitrogen instead
of air.  I'm not sure of that, though, and if preservation isn't
perfect sequencing could become a problem without making
identification impossible.  DNA is terribly sensitive to all
kinds of damage, and enzymes already present in the blood or
tissue will tear it up given half a chance.

    Re genomic analysis: yes, it's certainly true that DNA
sequencing is doable at the moment on the scales the human genome
would require, in the same sense that space flight was doable in
the fifties.  It's logical to predict that it will only get
easier as automatic sequencers get better.  The closest tome I
happened to grab quotes the length of the human genome at about
2.9 x 10^9 base pairs.  The fact that there are four possible
bases (2 bits) gives you a 5.8 billion bit storage issue.  Not
that intractable for storage and analysis, especially given that
some compression technques that wouldn't work well for most data
would be applicable.

James Hicks comments -
>"Single Cell" polymerase chain reaction (PCR) is being done in the lab now.
>Theoretically all you need is one cell and you can amplify any DNA
>sequence from the genome that you want.

    PCR makes tiny sample sizes a lot less of a problem than they
used to be, but it has the same problems any extremely sensitive
amplifier does.  It amplifies everything.  If there's the least
contamination of the sample with any other DNA, the analyst is
in trouble.  Suppose you vaccuum a chair.  You get some skin from
me, some skin from N other people, umpteen dust mites and the
foot of a crushed roach.  Given the way the enzymes in the dead
cells would have torn up the DNA, you may get nothing but if
you get anything, the bugs win.  Research labs have had terrible
trouble with contamination - some PCR amplified "human" DNA
in the big databases turns out to look suspiciously like yeast.

and //mb adds -
>the streets for saliva every morning at 3am and whipping the flesh of all
>offenders.

    Saliva would give the same problem.  Nobody's mouth is sterile,
and my normal bacterial flora is a lot better protected against
the digestive enzymes in saliva than shed cells from my mouth are.

    Given all that, if anyone is still awake, it's the step
*after* all the sequencing that's the biggie... at least for
anything beyond simple ID.  You've got a sequence: what does it
do?  A lot of the time, nothing.  Lots of animal DNA doesn't ever
get used for anything obvious and seems to be along for the ride.
You have to distinguish live data from red herrings.  Then if
you're looking for genetic predictors of disease, you can't
just say that *any* change in a particular gene is a red flag --
there's  a lot of function-neutral variation.  You'd be denying
insurance coverage to very safe risks and losing money.  But
when a change is *not* function-neutral, it may only take one
base-pair change. Sickle-cell anemia is produced by just one "typo".

    What makes it even harder is that most genetic
predispositions to disease probably aren't single, consistent,
easy to spot changes.  A lot of the ones we know about are,
but only because those are the ones it's easy to find.
Considering that interaction effects really aren't well studied
even in pharmacology where they've been known longer (What
happens when somebody mixes prozac with alcohol and marijuana?
The last time I checked Medline nobody had looked.) I think it
will take a long time to sort out problems that have something
to do with several genes plus an environmental trigger.

    The problem may not be big enough to be formally called
intractable, in the cryptographic sense, particularly if one
makes the customary (sensible) assumptions about processing
power increases, but it still looks big enough to be interesting.

    Sequencing is necessary for some of the 1984ish outcomes
predicted, but not sufficient.  Conversely you can do a lot of
unpleasant discriminatory things to people on the insurance front
without knowing their DNA sequence -- Down's Syndrome is
extremely obvious and a clear indicator of a bunch of expensive
problems not to mention an early death.

    It looks to me like the issue is worth keeping an eye on,
but contagious diseases in the waiting room are still a better
justification for avoiding the medical profession than a DNA
registry is.
                   regards...

                                      -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Dept. of Biology                             Jennifer Mansfield-Jones
University of Michigan                             cardtris@umich.edu