Hi Scott
yes, I agree that defining sets of genes that we can hypothesise were
present in the last common ancestor is a very valuable task. Indeed
weve been doing it for nematodes, and its a fun and revealing game to
get into, however done (domains/genes/etc). But I thought from Don's
post that he was commenting on the best blast matches (vide "the best
gene matches to human genes"), not presence/absence of members of
inferred orthologue groups present in the LCA of humans and
arthropods. Thus Daphnia, for human gene X, might 'win' the race, but
this doesnt imply that Apis/Ixodes/Cupiennius DOESNT have a copy of
the family X, just that whatever copies were present scored less
highly than Daphnia's.
Mark
On 1 Feb 2010, at 21:42, scott cornman wrote:
> Hi Mark,
>> I agree with you that BLAST is not a proper method for inferring
> phylogenies, but are tree topologies even needed to formulate sets of
> ancestral genes? Perhaps I'm being naive, but can one just use a BLAST
> threshold to estimate whether homologous genes are present/absent in a
> clade, and then identify the arthropod with the most overlap with
> vertebrates but not necessarily the most similar sequence by some
> evolutionary model? And given the rate of multigene family evolution,
> perhaps presence/absence of well defined domains is a better unit to
> tabulate?
>> Scott Cornman
> Beltsville Bee Research lab
> USDA-ARS
>> On Mon, Feb 1, 2010 at 3:59 PM, Mark Blaxter <mark.blaxter from ed.ac.uk>
> wrote:
>> Hi folks
>> I'm always worried when BLAST is used as a phylogenetic tool...
>> basically
>> the 'scores' returned are contingent on the underlying model, which
>> in BLAST
>> is evolutionary/biological of course, but is generalised across all
>> proteins. Statements of 'basal' imply a tree, and a tree implies a
>> model
>> driven reconstruction, and, well, BLAST doesnt reconstruct trees.
>>>> A top blast match (or reciprocal best blast) algorithm is severely
>> biased by
>> - the relative rates of substitution in each genome
>> all extant genomes are equally 'evolved'; the differences
>> are in
>> rates and patterns of rates of evolution across genes
>> two genomes with relatively low substitution rates will
>> appear more
>> similar than expected when compared to their sister taxa, when
>> sister taxa
>> have elevated rates
>> - the patterns of substitution that are driven by other forces,
>> such as base
>> composition
>> its well known in mitochondrial genomes that protein
>> substitution
>> patterns are driven by genome AT content, resulting in significant
>> and
>> completely confounding homoplastic convergence at the amino acid
>> level
>> thus global comparisons using BLAST, using a single
>> substitution
>> matrix (that implicitly assumes equivalence of amino acid
>> composition bias)
>> will mislead
>>>> It would be good to derive trees for all of the protein sets Don has
>> generated and ask what their topologies are. Are these genes not in
>> TreeFam,
>> where there are trees derived that can be mined for data affirming
>> this
>> finding.
>>>> Mark
>>>>>> On 1 Feb 2010, at 17:40, Don Gilbert wrote:
>>>>>>>> Scott,
>>>>>> Yes, to be sure, I qualified it 'the most ancestral arthropod gene
>>> set (of
>>> those available)'..
>>>>>> It may be a slog thru many arthropod genomes to find "the ancestral"
>>> model;
>>> from insect genomes, it is clear even the phylogenetically basal
>>> ones are
>>> more derived than ixodes or daphnia, and that agrees with known
>>> phylogeny
>>> (?). Part
>>> of Ixodes lower score is its highly repetitive genome, subsequent
>>> poor
>>> assembly, and
>>> artifactually poorer gene models (missing in all or parts). But the
>>> distinction
>>> in Ixodes better score to mammal genes vs plant/worm suggests a real
>>> difference
>>> from Daphnia.
>>>>>> Another thing I noticed in these stats of non-arthropod homology,
>>> I can't
>>> put too much
>>> weight on its biological accuracy but someone could investigate
>>> further:
>>> pediculus (mammal parasite) and aphid (plant parasite) are two
>>> relatively
>>> close insects,
>>> and should by phylogeny have about same gene distance from non-
>>> arthropods.
>>> Not quite:
>>> pediculus favors mammal genes over aphid by 2%, and aphid favors
>>> plant
>>> genes over
>>> pediculus by 1%, the only case I looked at where aphid exceeds
>>> louse.
>>> More details are here:
>>>>>>http://arthropods.eugenes.org/arthropods/summaries/best-arp-model6genes.txt>>>>>> - Don
>>>>>> |Subject: Re: Daphnia has the most ancestral genes
>>> |
>>> |Hi Don,
>>> |
>>> |I thought that was an interesting comparison you posted on
>>> Arthropod
>>> |Digest, but I wonder if there will be a non-parasitic arthropod
>>> genome
>>> |coming out that would be a better indicator of ancestral genes
>>> (eg a
>>> |spider?). Mites and ticks have a long history of parasitism and
>>> that
>>> |has been shown to be associated with gene loss, I believe.
>>> |
>>> |Scott Cornman
>>> |Beltsville Bee Research Lab
>>> |USDA-ARS
>>> |
>>>>>> _______________________________________________
>>> Arthropod mailing list
>>>Arthropod from net.bio.net>>>http://www.bio.net/biomail/listinfo/arthropod>>>>>>>>> --
>> The University of Edinburgh is a charitable body, registered in
>> Scotland, with registration number SC005336.
>>>>>
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
