[BioC] Questions about Disease Progression Analysis

[Sean Davis]

On Mon, May 19, 2008 at 10:08 PM, LiGang <luzifer.li at gmail.com> wrote:
> Sean Davis <sdavis2 at ...> writes:
>
>>
>> On Mon, May 19, 2008 at 5:00 AM, LiGang <luzifer.li at ...> wrote:
>> > Dear all,
>> >
>> >
>> >
>> > I am using cDNA two-channel array to study gene profiling of 5 different
>> > stages of thyroid disease. Was there any Bioconductor package to perform
>> > such disease progression analysis?
>> >
>> >
>> >
>> > I have found packages such as
>> > "maSigPro<http://bioconductor.org/packages/2.2/bioc/html/maSigPro.html>
>> > "  "Mfuzz <http://bioconductor.org/packages/2.2/bioc/html/Mfuzz.html>" and"
>> > timecourse
> <http://bioconductor.org/packages/2.2/bioc/html/timecourse.html>"
>> > to perform Microarray Time Course Data analysis, can these methods be used
>> > to carry out disease progression data analysis?
>>
>> Hi, LiGang.  There is a real temptation to think of disease stage as a
>> process that occurs in an ordered fashion.  That is, stage I disease
>> is just early enough that it has not progressed to stage II, etc.  I
>> think that there is plenty of evidence that this not always (or even
>> often) the case, so I would be hesitant to treat the stages of thyroid
>> disease as a disease progression.
>>
>> As for time course analysis, it usually examines the behavior of genes
>> in the same sample(s) over time; you will likely not have the same
>> person who has multiple stages of thyroid disease, so I am not sure
>> that these methods will be applicable in your situation, anyway.
>>
>> Do you have clinical followup data?  Other clinical covariates?  And
>> how many samples do you have in your dataset?  All of these are
>> important questions that can guide the hypotheses that you might want
>> to test.
>>
>> Hope that helps.
>>
>> Sean
>>
>> _______________________________________________
>> Bioconductor mailing list
>> Bioconductor at ...
>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>> Search the archives:
> http://news.gmane.org/gmane.science.biology.informatics.conductor
>>
>>
>
> Dear Sean,
>
> Thanks for your reply!
>
> In fact, there are no clinical parameters except the qualitative stage
> information.
>
> Experiment details are listed below:
>
> Chip    Stage   sample
> ================================
> Chip_1  Stage 1 mouse_1_thyroid
> Chip_2  Stage 1 mouse_2_thyroid
> Chip_3  Stage 1 mouse_3_thyroid
> Chip_4  Stage 1 mouse_4_thyroid
> ----------------------------------------------------
> Chip_5  Stage 2 mouse_5_thyroid
> Chip_6  Stage 2 mouse_6_thyroid
> Chip_7  Stage 2 mouse_7_thyroid
> Chip_8  Stage 2 mouse_8_thyroid
> ----------------------------------------------------
> Chip_9  Stage 3 mouse_9_thyroid
> Chip_10 Stage 3 mouse_10_thyroid
> Chip_11 Stage 3 mouse_11_thyroid
> Chip_12 Stage 3 mouse_12_thyroid
> ----------------------------------------------------
> Chip_13 Stage 4 mouse_13_thyroid
> Chip_14 Stage 4 mouse_14_thyroid
> Chip_15 Stage 4 mouse_15_thyroid
> Chip_16 Stage 4 mouse_16_thyroid
> ----------------------------------------------------
> Chip_17 Stage 5 mouse_17_thyroid
> Chip_18 Stage 5 mouse_18_thyroid
> Chip_19 Stage 5 mouse_19_thyroid
> Chip_20 Stage 5 mouse_20_thyroid
> ================================
>
> and all 20 hybridizations use the same common reference sample (pooled samples
> of normal tissues from the above 20 mice).
>
> My aim is to identify genes that are up-regulated or down-regulated as thyroid
> disease progresses and even whether there exist genes which have a specific
> trend.
>
> Is it rational to:
>
> 1)      perform pairwise comparisons
> 2)      Select the Union set of all genes whose expression level changed
> between any arbitrary pairwise comparisons.
> 3)      Do cluster analysis to see the trends.

See section 8.6 of the limma user guide on multiple groups as an
example of how an analysis of multiple groups could be done.

Sean