In article <4b74s5$57u at news.nstn.ca>, "Andrew J. Roger" <aroger at ac.dal.ca> writes:
>carmean at sfu.ca (Dave Carmean) wrote:
>>In article <DJK4In.ELC at zoo.toronto.edu>, mes at zoo.toronto.edu (Mark
>>Siddall) wrote: [much interesting and useful material here and in related
>>posts]
>>>>> However because MALIGN is so thorough it takes a lot of memory and
>>> processor time.
>>> So you have a choice... get your alignment quick or be rigorous in your
>>> science.
>>>>I find it interesting that there is no discussion here of using homology
>>(such as secondary structure) to align sequences. Karl Kjer has been
>>working on this aspect of alignment.
>> There is an assumption here however- that is that secondary structures
> will change more slowly than primary sequences. I think that this
> is generally true but one should look out for cases where there
> are secondary structures which appear identical but are shifted
> relative to primary sequence similarity. In other words, it is
> possible that structures could be retained by shifting over several
> bases causing you to misalign the primary sequences. One would
> expect, for instance, that sometimes in stemloops that loops
> could get one base smaller (by a new base-pairing at the top of
> the stem) and the base of the stem could unpair. You would be left
> with a stem of the same size but its alignment with homologous
> stems would become problematic if stem boundaries are only paid
> attention to.
>> It worries me when I here many rRNA advocates claim that they
> have based their alignment on secondary structure and therefore
> homologous alignment is assured. Homology is ALWAYS an inference-
> and nothing guarantees homology- not even tertiary structures!
>> Cheers
> Andrew J. Roger
>
That is an interesting point Andrew but is only a conjectured problem.
It could occur but I do not know of any examples; this does not prove
it cannot happen, just that it is likely to be a rare phenomenon if it does.
You can look at cartoons of secondary structure and say things like "...oh
imagine a loop extends here .... and a loop shortens there" but this
does not neccessarily have any biochemical reality. You cannot just move
stems around willy nilly just because you get something that "looks like"
what you started with. There is MUCH more to rRNA structure and
function than having a stem of roughly the right size more or less
in a certain place. There are now enough
examples of sequences of rRNA to spot this happening if it happened
very often.
When you align small subunit rRNA sequences, you get large conserved
blocks with high sequence identity and very few or no gaps (the conserved
core) seperated by horrible bits that are full of gaps and almost impossible
to align (with or without reference to secondary structure). You can avoid
this conjectured problem by only using the bits that can be unambiguously
aligned. For some difficult bits, there might not be ANY meaningful alignment.
This does not PROVE there are no problems but I would love to see some real
examples.
Des Higgins
>
