[Taxacom] SINES, LINES and chromosomal rearrangements

[Pierre Deleporte]

Le 09/05/2011 16:51, John Grehan wrote :
> -----Original Message-----
> From: taxacom-bounces at mailman.nhm.ku.edu [mailto:taxacom-bounces at mailman.nhm.ku.edu] On Behalf Of Pierre Deleporte
> Sent: Monday, May 09, 2011 10:09 AM
> To: taxacom at mailman.nhm.ku.edu
> Subject: Re: [Taxacom] SINES, LINES and chromosomal rearrangements
>
>> OK, my statement wa a shortcoming
>> the fact is that your ingroup is so small that your approach _nearly
>> boils down to_ a priori clique analysis
>> because you reject from the analysis characters showing some homoplasy
>> in ougroups
> You mean I reject from the analysis of the ingroup relationships those characters that show the same character state in the outgroup (in this case gibbons and OW monkeys)?

exactly,
would you perform an analysis at a larger scale, including gibbons, you 
would use these characters, isn't it ?

so it appears that for you, some characters would be good when you 
perform a large-scale analysis, and bad when you ar working on a smaller 
ingroup

are you aware of such contradictions in your approach?
>> if you included more species in your ingroup, some characters you reject
>> a priori because you already observed their homoplasy would be included
>> in the analysis,
>> would you use your usual standard parsimony analysis with PAUP (and, OK,
>> not a clique analysis)
> But I included all species that are known to exist. What else can one do?
vary the size of your ingroup and observe that your character set / data 
matrix changes according to such choices, according to your own 
professed method (= rejecting characters from the ingroup when they 
happen to show some homoplasy in distant outgroups)
or, more simply, renounce to "uniquely" derived (see below)

>> now your position did not change:
>> you will not analyse molecular data
>> although you could perfectly analyse them "your own pet way"-
>> this may be a comfortable psychological protection (refusing to see what
>> happens),
>> but it has no logical justification
> To the contrary - there is a logical justification, but of course one over which we differ.
this cannot be, logically
all scientists use the same logical rules = we all agree on mathematics

example: if you enlarge your ingroup, some previously rejected 
characters will become useful ones
this is pure, universally accepted  logic, and no scientist can escape it
so I am logically justified to ask you why you are using such a method, 
with such internal contradiction

unless you would have a logical - and biological - argument justifying 
the idea that a biological character would be either phylogenetically 
informative or flawed according to the size of the ingroup you decided 
to analyse
my logical and biological point is that the way a researcher decides to 
delimit the range of an ingroup at stake should not change the phylogeny 
of the lineage under scrutiny: your arbitrary decision cannot change the 
evolutionary history
hence a good caution should be to vary the size of your ingroup - and 
what would you do with the contradictions...
but you can dispense with this if you renounce to the criterion "unikely 
derived characters" (typical of clique analysis logic: every clique is 
composed of non-homoplastic and mutually compatible characters = 
uniquely derived in your own terms) and adopt "derived characters" = the 
"PAUP" logic (standard parsimony) you already used

just stating "we differ" without analysing the divergence it is typical 
of post-modern radical relativism "anybody his own truth"
this cannot be the way of science: we are supposed to argue and progress 
toward mutual logical-biological agreement
this requires to accept all the logical points of the opponent, the 
argument "our logics differ" being logically excluded
such recommendations are part of basic courses on the scientific method

>> so you go on saying that molecularists analyse their data the wrong way
>> = "phenetically",
>> even when they analyse them cladistically (e.g. parsimony with PAUP,
>> just like you...)
> Yes they use clustering algorithms used by cladists, but it is my contention hat this does not make their analysis cladistic because they cannot (or have not so far) restrict the character data to shared derived states.

yes they could, but they don't do it notably because of logical 
contradictions as mentioned above
so do it yourself and see what happens - what do you fear?
it is very simple :
reject your bad molecular characters, retain your good ones (no 
homoplasy in outgroups) and perform the standard cladistic analysis

it is highly unfair of yours to refuse to save contemporaneous 
phylogenetic molecular analysis from its present state of disaster
not to mention the fact that re-analysing all the recent phylogenetic 
data sets "the right way" opens you a highway for numerous high-standard 
publications
you just have to convince your reviewers that all current molecular data 
are phenetic but will become cladistic when you will have purged them 
from characters (potentially) non-uniquely derived in the ingroup you 
delineated - and then use, for analysis inside the ingroup, an algorithm 
that does not optimize according to "unique derivation" (your previous 
criterion), but according to standard "derivation" (= optimizing 
homology, not forbidding homoplasy like you did for constituting tour 
data matrix)

you are cliquist outside (no homoplasy), and standard cladist inside 
(minimize homoplasy), and your ingroup delineation is arbitrary - I call 
it logical incoherence, with no offense to the man, just discussing logics

>> and alternatively / complementarily you qualify molecular data of being
>> "phenetic" in themselves,
>> which make no sense at all under any definition of "phenetics" since the
>> beginnings of phenetics
> In phenetic analysis there is no restriction of character states to those that are shared derived.

everybody should know since Aristotle that a definition cannot be based 
on the absence of some property
please define phenetic analysis positively, based on the properties of 
the analysis (not what it is not)

and the data set is not the analysis of the data set, so your definition 
is insufficient
an "exhaustive" data set is still not a data set analysed phenetically,
it is just a raw exhaustive data set
so, deciding to use this complete data set still not defines phenetics 
(= clusteting on the basis of overall similarity rather than e.g. a 
minimum length cladogram optimising character state contiguity using a 
Mahattan distance)

in other words, you still could analyse your own restricted data set 
phenetically
pure logics...

>> now 3-item analysis pops up in your last post below, really at no
>> surprise for me
>> a must reading about this bizarre approach is Farris 2010, "Systematic
>> foundering", Cladistics 27 p.207-220
>> ( = a criticism of the recent book by Williams and Ebach "Foundations of
>> systematics and biogeography")
>> 3-item analysis is not the standard parsimony analysis (neither
>> classical "cladistics", nor clique analysis)
>> it simply does not optimize character state contiguity on the cladogram
>> this approach has never been the object of any attempt at a biological
>> (or biogeographical) justification,
>> particularly not at any evolutionary justification
> Well, each to their own as far as what is bizarre or not. I liked the idea of each character being represented as a definitive relationship.
this is not a definition of 3-item analysis
this was also told of standard parsimony / "cladistics" = representing 
"relastionships" some way (see "pattern cladistics")

and more than this : could you define "relationship"? and after that, 
"definitive relationship"?
(a non-tautological way of course, not "definitive relationship is what 
3-item analysis reveals")

>> among other bizarre statements, the authors consider that molecular data
>> have no hierarchical structure,
>> which sounds much like your own position that molecular data are
>> "phenetic" in themselves
> Well, whether it is bizarre or not remains to be seen. No more bizarre than the contention that morphology is unreliable and cannot provide the foundation for the tree of life while at the other side of the mouth they say that fossils can be definitively recognized in their relationship to the living and are so definitive that they can be used to calibrate molecular clocks.
>
it is not because other people are sometimes wrong that you are always 
right (logics once more)
and their errors don't allow you to commit similar mistakes (logics again)

but these points are completely outside our present debate - I see it as 
a possible rhetorical diversion = throwing balls in all directions, 
perhaps expecting that I would run after them - the more when I never 
said that morphology is not reliable, to the contrary I am co-authoring 
pluri-data analyses (notably showing that behavioral data can perform 
pretty well in face of molecular and morphological ones)

now my point is that the argument that molecular data are by nature 
phenetic or non-hierarchical, hence unuseful for phylogenetic analysis, 
is bizarre in itself: "data phenetic by nature" makes no sense


>> now the authors also consider that standard "cladistic" parsimony analysis
>> (like you performed with PAUP) _is "phenetic"_
>> more generally, it appears that any non-3-itemic approach is "phenetic"
>> (according to an odd, esoteric concept of "pheneticism" of course,
>> distorted from a comment by Nelson in the eighties...
>> but it is a too long and boring story of nonsense to tell)
> I can see that PAUP of itself is just a similarity method for clustering data and there have been enough papers out there that draw attention to those constraints. But as I said, I use it because everyone else does and it suits me to do so for that reason.

do you know of any clustering or classificatory method that does not 
implement some "similarity" criterion, in the largest sense of the term? 
I don't know of those "enough papers out there" (except possibly from 
the peculiar 3-itemist school of course)

or do you precisely mean "overall similarity"? are you also saying that 
parsimony (standard cladistic) analysis is "phenetic"?
if you mean that parsimony analysis clusters on the basis of an index of 
overall similarity, this is wrong (pure logics)

standard parsimony analysis clusters by shared synapomorphies, but not 
"uniquely" derived ones (this is clique analysis, and still not any sort 
of phenetics)
and it uses a contiguity criterion for homology, because characters are 
supposed to be (generally) phylogenetically inherited from generation to 
generation by descent with possible modification (is it what bothers 
you, like Ebach and Williams ?...)

>> good luck John if you try and understand this one...
> And good luck to you Pierre in your efforts to understand different perspectives.

decifering logical incoherences is not a matter of luck
yes I like to try and understand different points of view, but a 
radically anti-postmodern way:
I want to make up my mind about what makes sense and what doesn't,
and debate for promoting mutual agreement - logically, and biologically

3-item analysis is formally coherent (just so), but devoided of any 
biological justification
(I took the pain to write about this, directly comparing standard and 
3-items logics, and Jan De Laet did even a better job , in showing that 
the basic difference consists in the indifference of 3-it regarding 
character contiguity/ historical continuity;
the elementary "unit of information" in standard cladistics is a 2-item 
statement of character state contiguity + outgroup rooting for polarity, 
not a 3-item statement of relative inclusion/exclusion in clades with no 
contiguity constraint;
standard contiguity makes phylogenetic sense, while 3-itemists never 
justified their method in biological terms)

"cladistic" analysis is not phenetic, while "similarity" is so vague 
that it practically qualifies all classificatory approaches

JG sometimes uses cliquist criteria in parsimony analysis, sometimes 
not, calling this mixed approach "cladistic",
JG argues that all data should be used, not only molecules, while he 
refuses to analyse molecular data "the right way";

yes I object on (universal) logical grounds

Pierre

-- 
Pierre DELEPORTE
UMR 6552 EthoS
Université Rennes 1, CNRS
Station Biologique
35380 Paimpont
tél (+33) 02 99 61 81 63
fax (+33) 02 99 61 81 88