[BioC] Further analysis after using MEDME?

[mattia pelizzola]

Dear Prashantha,

for example you can select probes with significant methylation level
in at least one sample.
If AMSmat is the matrix with your AMS values for the four samples, you
can do that filtering based on the max value for each probe, like:

maxs=apply(AMSmat, 1, max, na.rm=T)
dataF=AMSmat[maxs>thr, ]

where thr is an adjustable parameter

otherwise, if you have a baseline sample the other samples can be
referred to, you can determine differential methylation in respect to
this sample. Let's suppose the sample in column 1 is a baseline for
all the others.
Again you can then select only the probes with at least a given thr
methylation log2Ratio (like thr=1.58 for Ratio >3). Note the use of
the abs function to select at the same time for 3 fold hyper- and
hypo-methylation:

diffMetMat=log2(AMSmat[,2:ncol(AMSmat)] / AMSmat[,1])
maxs=apply(abs(diffMetMat), 1, max, na.rm=T)
dataF=diffMetMat[maxs>thr, ]

at this point you selected the probes with high methylation level (or
relative methylation level) in at least one sample. Now you can use
the heatmap.2 function to get an heat map of those. Note that if you
still have a dataset with several thousands probes the clustering will
take a while and require a good amount of memory:

require(marray)
require(gplots)

# for the first analysis scenario:
cols=maPalette(low='beige', mid='gray', high='red', k=128)
Breaks = seq(1, 32, length.out=129)

# for the 2nd analysis scenario:
cols=maPalette(low='blue', mid='beige', high='red', k=128)
Breaks = seq(-1.58, 1.58, length.out=129)

png('test.png', 1000, 1500, res=200) # to save the heatmap on a test.png file
heatmap.2(x=dataF,breaks=Breaks, Colv=FALSE, Rowv=TRUE,
dendrogram="none", scale="none", col=cols, density.info="none",
symkey=F, trace="none", margins=c(5,10), labRow=FALSE,  labCol=FALSE)
dev.off() # to save the heatmap on a test.png file


best,

mattia


On Fri, May 15, 2009 at 5:36 AM, Prashantha Hebbar Kiradi [MU-MLSC]
<prashantha.hebbar at manipal.edu> wrote:
>
> Dear Dr. Plizzola,
>
> Now I am done with artificial fully methylated data and MEDME analysis. So,
> I would like to show you the data and ask doubts to go further.
>
> Following is the data obtained from smooth function
> "V1"    "V2"    "V3"    "V4"
> "A_17_P16499695"        0.6494378084742 -0.492466374892136
> -0.422196325432134      -0.0320280553522771
> "A_17_P00917694"        -0.203727237011732      0.0180055762341851
> 0.00673134043866919     0.437505835919691
> "A_17_P05822757"        -0.2624830602918        -0.267402421470106
> -0.892633555689626      -0.095066617651747
> "A_17_P11201690"        0.398597328778726       -0.115790664303932
> -0.50134339502071       0.0372322294924269
> "A_17_P15518473"        0.256514445796715       0.443915215417359
> 0.0364180402923375      0.0776241952017851
> "A_17_P11189908"        0.718052106295435       0.0492163040883114
> -0.770470136219705      -0.241915478429616
> "A_17_P17268257"        0.693399037168038       0.608191021500769
> 1.11052034793374        1.21791753773972
> "A_17_P07299116"        -0.454641426498295      -0.0108170868279176
> -0.603155929394701      -0.147985922703689
> "A_17_P10258961"        0.221998685136869       0.105838604181660
> -0.609621004305267      0.275607454708045
>
>
> Following is the data obtained from AMS function
>
> "V1"    "V2"    "V3"    "V4"
> "A_17_P16499695"        32      1       1       32
> "A_17_P00917694"        1       32      32      32
> "A_17_P05822757"        1       1       1       29.8234198110255
> "A_17_P11201690"        32      27.0622296124447        1       32
> "A_17_P15518473"        32      32      32      32
> "A_17_P11189908"        32      32      1       1
> "A_17_P17268257"        32      32      32      32
> "A_17_P07299116"        1       32      1       17.2568692096704
>
>
> Following is the data obtained from RMS function
>
> "V1"    "V2"    "V3"    "V4"
> "A_17_P16499695"        0.630019152582239       0.0224006809807018
> 0.0224006809807018      0.525084096217918
> "A_17_P00917694"        0.40485628843651        0.558850704897022
> 0.635847913127279       0.635847913127279
> "A_17_P05822757"        0.0244329519078063      0.0244329519078063
> 0.0244329519078063      0.141806490251356
> "A_17_P11201690"        0.719610510811024       0.531449873540947
> 0.371049169636934       0.680448534158954
> "A_17_P15518473"        0.479067027522034       0.580720092915215
> 0.573913209854228       0.355691056910569
> "A_17_P11189908"        1.31194857161599        1.24137524424337
> 0.0409983928629998      0.0409983928629998
> "A_17_P17268257"        0.552604984195927       0.619261616530845
> 0.619261616530845       0.619261616530845
> "A_17_P07299116"        0.0315159155373464      0.437289500615857
> 0.120181991903205       0.245333995626829
>
>
> Now I want to continue this for further analysis like to fetch Hyper and
> Hypomethylated probes. I.e., how should I make filtration? For example,  In
> "A_17_P05822757"        1       1       1       29.8234198110255 (In the
> above data "V1" corresponds to artificially fully methylated data) in order
> to say, Probe A_17_P05822757 is Hyper / Hypo methylated Should I compare the
> values across the experiments (Ex, V1, V2, V3, V4) and decide manually? I
> also would love to get heatmap for the filtered data. I tried in many ways
> but I could not. Only thing on which I am worried is how to bring the data
> in presentable form? So, can you please suggest me how should I go further?
>
> Thanking you in anticipation.
>
> Regards,
>
> Prashantha,
> Bioinformatician,
> Manipal University,
> India
>
> ________________________________
> This e-mail is privileged and confidential. If you are not the
> intended recipient please delete the message and notify the sender.
> Any views or opinions presented are solely those of the author.
> ________________________________