Jack Campin wrote:
>Here's a set of tunes that for a dead cert are genetically related.
>Do any of the tune-matching algorithms suggested here detect that?
I converted each of them to a list of intervals, and assigned each
interval to a single character IUB protein code. Since there are
only 20 of these, I assigned any intervals outside the range
+-9 semitones to a single catchall code.
I then performed a multiple alignment on these "proteins" using
the Clustal algorithm. This tries to find the optimal arrangement
of the sequences so that you can see areas of similarity where
the symbols line up. The results are probably far from optimal,
and a better choice of symbol for each interval would certainly
improve things, but nonetheless interesting:
10 20 30 40 50
| | | | |
| ||*
Mary Scott LVLLILTIILLVLILIL---------IDLVLLILT------IIRLI------
When Cold LL-VKV----LVNE------LLSQI---LILVYLLVKVLVNIIRLI------
Untitled LLSLKV----LVNIILLVLVLLSGI---IKIDLLSLKVLVNIIRLI------
Sir John Fenwick LLSLLVTLVKIIGIILLSLLAVKAVKIDLLSLLVTLVKIIGIIRLI------
Carraig's Rant QVAAALILVTKAAAELLILRAAACSGIIKILVYQVAAALILVTKIIRQEQTV
Fireman AQVAAKVLVTK----IILAVQAAIRQIIKA-VYAQVAAKVLVTKIIRQEQTV
60 70 80 90 100
| | | | |
** |* | |*
Mary Scott -GILSGIILLVLIRAGII-----LVNIIYAILLII-LIGS------AGIILV
When Cold -GILSGIILLVLIRA-ALL---PLVNELY--AAIL-LSGII--KIRAALLPL
Untitled -GILSGIILLVLIRYYALL-IITLVNIIYYYLAIL-LSGII--KILSLYYAL
Sir John Fenwick -GILSGIILLVLIRYQHYQHYYKIIGIIYYVFYVF-YYTIRYQHYQHYYKII
Carraig's Rant KVTLSLLILVALIQEAAATLIRTKAAAEYLILIAAALSGIIKAAAIRQEAAA
Fireman KVTLSLLIRLLII---AATLIRTK-----IIYALIALSGIIK---IRLLIIA
110 120
| |
| *|*|
Mary Scott NIIRLIGILSG----------IILLVLI
When Cold VNERLIGILSG----------IILLVLI
Untitled LIITLVNIIYSGIIK----VKIIKLVLI
Sir John Fenwick GIIRLIGILSG----------IILLVLI
Carraig's Rant TLIRTKAAAERAAAQTAAAVTLSLLILV
Fireman TLIRTKIIRQEQTVK----VTLSLLI
Also interesting is the multiple alignment log. Clustal starts
by performing a pairwise alignment for every possible pair of
sequences, and it gives you a score for each. For proteins,
you expect a score of 100 for identical sequences, and about
5 for completely unrelated sequences:
Aligning Mary Scott with :
When Cold Score: 42.05
Untitled Score: 40.91
Sir John Fenwick Score: 42.05
Carraig's Rant Score: 19.32
Fireman Score: 23.86
Aligning When Cold with :
Untitled Score: 46.74
Sir John Fenwick Score: 34.78
Carraig's Rant Score: 23.91
Fireman Score: 22.83
Aligning Untitled with :
Sir John Fenwick Score: 34.55
Carraig's Rant Score: 17.27
Fireman Score: 21.82
Aligning Sir John Fenwick with :
Carraig's Rant Score: 15.79
Fireman Score: 16.36
Aligning Carraig's Rant with :
Fireman Score: 32.73
As you can see, the first four tunes show strong similarity
with each other. The last two form a separate group and
are more strongly related to each other than to the others,
although they all clearly are related.
Even with this relatively crude conversion you should have
no trouble pulling these tunes out of a database by searching
with any of them.
If I get the time I'll try them on some of the phylogenetic
analysis programs tomorrow, and see if I can produce a
family tree.
Phil Taylor
To subscribe/unsubscribe, point your browser to: http://www.tullochgorm.com/lists.html
