Re: Article: On Noise in Gene Expression

"Malcolm" <regniztar@xxxxxxxxxxxxxx> wrote in
news:dm0r0a$1fld$1@xxxxxxxxxxxxxxxxxxx:

>
> "Perplexed in Peoria" <jimmenegay@xxxxxxxxxxxxx> wrote
>>
>>> In fact the datasets are very difficult to work with. For instance I
>>> found a
>>> set of four genes (out of 2000) which provided a very statistically
>>> significant discrimination between tumour and non-tumour samples,
>>> but no sort of rhyme or reason or real biological plausibility.
>>
>> I'm not an expert, but it would seem that the next step would be to
>> move from two gene samples (tumour and non-tumour) to a sample from
>> every cell type (embryonic or adult) in the organism. You need an
>> array (or rather a tree) of microarrays.
>>
> The situation is that I can distinguish a tumour from a non-tumour
> sample with a probability much greater than chance, but not good
> enough to be of any practical use (eg in a path lab). What I cannot do
> is provide any sort of biological justification for the rules my
> program uses to distinguish the two sets.
> Adding more tissue types may illuminate, but I suspect it would just
> pile on confusion. If I cannot do a simple binary, it is unlikely that
> the program will work on a harder problem.

If I am following you, you have found a set of four genes which, if
expressed in a given cell are a good indication that the cell is
cancerous. There are too many false positives and false negatives for the
test to be useful in diagnosis. But you say there is no "biological
justification". I am assuming your program looked for a statistical
association but that the known functions of the genes in question are not
associated with the known pathways for creating cancerous cells from
normal cells.

While correlation is not causation, couldn't your results still be a
useful lead to follow in teasing out the steps that can make a particular
cell cancerous, especially in determining why some individuals get cancer
while others with otherwise similar genetics and histories don't ?

Yours,

Bill Morse

.

Relevant Pages

  • Re: Minimum Cell?
    ... What must the cell be able to do? ... Note that mitochondria have only a half-dozen genes related to ... It is generally believed that hydrogenosomes are ...
    (talk.origins)
  • Re: epigenetics
    ... production of a pine tree but when put into a human egg cell would ... PAX gene into leg cells of a fly causes the formation eyes in the leg. ... It's because of the other genes ... "responsible" for producing eyes of whatever type. ...
    (talk.origins)
  • Re: epigenetics
    ... production of a pine tree but when put into a human egg cell would ... PAX gene into leg cells of a fly causes the formation eyes in the leg. ... It's because of the other genes ... "responsible" for producing eyes of whatever type. ...
    (talk.origins)
  • Re: Where does information come from?
    ... in the cell. ... describe the sequences of bases in DNA. ... You can describe blood vessels as pipes. ... that genes cause the expression of certain molecular building blocks, ...
    (talk.origins)
  • Re: Article: On Noise in Gene Expression
    ... > expressed in a given cell are a good indication that the cell is ... > association but that the known functions of the genes in question are not ... > cell cancerous, especially in determining why some individuals get cancer ... difficulty is compounded when you youself are not really a genetics person. ...
    (sci.bio.evolution)