Molecules et al

[Susan Farmer]

>You could feed the sequences into your calculator and do an instant
>parsimony analysis. At 1 byte per cell of the data matrix, and about 4
>times as many bytes per tree as there are sequences, you could store 30
>sequences of 800 bases, and 50 trees resulting from the analysis, in about
>30k. Figures are based on my own programming work, but with 4-8 times more
>packing of the data matrix to keep the storage down, at the price of
>slight slowness. You could also pack the trees still tighter if you wanted
>to, with no speed penalty. The source text of the whole parsimony program
>currently takes about 150k, but assuming the sequencer knows how to
>behave, error checking could be cut, and let's fix the options to some
>reasonable value. Code space could then be reduced to less than 80k ...
>Allowing space for simple graphics as well, you only have to get a 64k
>upgrade for your machine! Well, unless you want your sequences aligned.
>Make it a 128k upgrade, why not.
>
>Has anyone ever done anything like this? Suddenly I am quite excited.
>Though anyone who has a sequencer can also afford a desktop computer ...
>and by the time sequencers hit the 9.95 low, calculators will come with
>2G RAM as standard. Oh well, the experiment might still be fun.
>

Ah to be in Edinburgh!  Beautiful city.  I don't know if you're a Star
Trek fan or not, but there is a prototype version of the Tricorder for
sale -- it does barometric, temperature, color spectrum, humidity, etc.
and it's either 500 or 600 dollars US.  Much of the technology is out
there -- memory's cheap, small and fast.  (my other passion is computers)
I have an old early Winchester (all Hard Disk Drives were Winchesters
in the old days -- stored a whoke 10 Kilobytes per ...) that is approximately
4 times the size and cost of a 2 GB drive.

The other thing that probably holds back most development of this sort
is that while the nameless parsimony program fits in about 150K, it's
probably written in some High-level programming language like C or
Pascal.  If it were to be written in some sort of Assmebler language
it would in all probability be about half the size or less.  But programming
in any assembler language is *not* for the meek -- think of aligning
sequences by hand ...  (At least that's how it used to be in the olden
days. :-)  Compiler-generated code tends to be "large" because of all
the functions etc. that it carries around that you may not need for
your particular application.  I see Great Possibilities for this type
of resource, but until you can convince Large Funding Agencies of the
wisdom of investing, you'll have people doing it on their own time and
with their own resources.  Just think how much computers have changed
in the last 25 years -- then try to extrapolate into the future.
In 25 years, we may very well have a "real" handleld gizmo that functions
just like a tricorder -- down to sequencing from a scan.  (ok, so
you'll probably have to grind up some leaf and put it into the
machine ...)  I don't think that it's too far off, though.

Susan, grad student
sfarmer at goldsword.com