limma analysis when only Cy5 was used
1
0
Entering edit mode
@0f31c4b4
Last seen 5 months ago
United States

Hi,

I have .gpr files from a 2-channel array (Phalanx Human OneArray) experiment in which only Cy5 was used. 6 arrays: n=3 each treatment and control. (Data was originally analyzed with Rosetta, but not going there.)

xxxxxxxxxxxxxxxxxxx

--This block of code runs fine, populating both the Red and Green columns with values (Red, Green, Red background, Green background, genes, etc).

RGuw <- read.maimages(targets, source="genepix")

RGw <- read.maimages(targets, source="genepix", wt.fun=wtflags(weight=0,cutoff = -50))

boxplot(data.frame(log2(RGw$Gb)),main="Green background")

boxplot(data.frame(log2(RGw$Rb)),main="Red background")

Control & empty probes are easily identifiable from gene IDs

xxxxxxxxxxxxxxxxxx

Not sure how to proceed as far as

** Background subtraction

** Within +/- across array normalization

** As far as running limma, should I extract the Cy5 values and use a simple 1-column design matrix or submit the entire object and somehow code Cy3 & Cy5 columns so Cy3 is ignored or cancels out?

Thanks very much for any advice.

Dan Schnell

Daniel.Schnell@cchmc.org

MicroarrayData limma Microarray • 1.4k views
ADD COMMENT
1
Entering edit mode

I don't entirely understand. From the code you give it is apparent that the arrays have both Cy3 and Cy5 channels, so something must have been hybridized to both channels. What do you mean then by "only the Cy5 was used"? Can you show us the targets frame showing which treatments/conditions have been hybridized to the two channels for each array? Choosing the appropriate analysis requires consideration of both channels.

ADD REPLY
0
Entering edit mode

Thank you very much for your reply Gordon. Pasted in below is the information that I have that indicates only the Cy5 dye was used--see DATA PROCESSING

I constructed a simple targets frame for the files as follows:

FileName sampID trt batch Pdate

H001-0607006461.gpr C_H001 Control 1 3/31/10

H002-0607006459.gpr C_H002 Control 1 3/31/10

H003-0607006460.gpr C_H003 Control 2 3/31/10

H004-0607006462.gpr T_H004 Treated 2 3/31/10

H005-0607006464.gpr T_H005 Treated 1 3/31/10

H006-0606025969.gpr T_H006 Treated 1 4/2/10

*** EXPERIMENT INFORMATION

GROWTH PROTOCOL: DMEM-F12 with 5% FBS + 1xPenstrep

TREATMENT PROTOCOL: Lentiviral infection of shRNA plasmids

EXTRACT PROTOCOL: Five million cells were pellted and RNA was isolated with Qiagen Rneasy kit

LABEL PROTOCOL: High-quality RNA (RIN=10) was converted to double-stranded cDNA and amplified using in vitro transcription that included amino-allyl UTP. The product ws conjugated with Cy5 NHS ester. hyb protocol Fragmented aRNA was hybridized at 42°c overnight using the HybBag mixing system with 1X OneArray® Hybridization Buffer (Phalanx Biotech), 0.01 mg/ml sheared salmon sperm DNA (Promega, Madison, WI, USA), at a concentration of 0.025 mg/ml labeled target. After hybridization, the arrays were washed according to the OneArray® protocol.

SCAN PROTOCOL: Raw intensity signals for each microarray were captured using a Molecular Devicess™ Axon 4100A scanner, measured using GenePixPro™ Software, and stored in GPR format.

DATA PROCESSING: 2 replicate samples were averaged. The F532 channel data in each raw data file should be ignored. Only Cy5 dye was used for the experiment. The data from all microarrays in each experimental set was then passed to Rosetta resolver (v.7.2.2.0.0) for analysis. Testing was performed by combining technical replicates and performing statistical analyses using Rosetta Resolver’s proprietary modeling techniques. Normalization algorithm -> Median scaling.

VALUE DEFINITION Averaged normalized signal intensity

ADD REPLY
0
Entering edit mode

Is this a public dataset from GEO, or a private dataset?

ADD REPLY
0
Entering edit mode

In short, I don't know.

ADD REPLY
0
Entering edit mode

Cross-posted to Biostars https://www.biostars.org/p/488271/

ADD REPLY
1
Entering edit mode
@gordon-smyth
Last seen 5 hours ago
WEHI, Melbourne, Australia

Thanks for providing the experimental information. I must say these Phalanx seem quite strange to me. I've never before seen anyone scan a green channel on a series of arrays but for it not to be a genuine channel. Anyway, you apparently just have work with the Cy5 channel.

It is possible to instruct read.maimages to read only the Cy5 channel, but it's probably easier to read both channels and throw the green away. Read the two color arrays:

RG <- read.maimages(targets, source="genepix")

Then create an EListRaw object to hold the single-channel data:

x <- new("EListRaw")
x$E <- RG$R
x$Eb <- RG$Rb
x$genes <- RG$genes

Then you proceed as for a single channel analysis.

y <- backgroundCorrect(x, method="normexp")
y <- normalizeBetweenArrays(y)

and so on.

Original answer

Two color arrays should be analyzed as two color arrays and I have never seen an example where it would be a good idea to extract one channel and pretend it was a single channel platform. On the other hand, limma does provide a pipeline allowing you to make separate channel comparisons from two color arrays if that is necessary. See the relevant chapter of the User's Guide.

ADD COMMENT
0
Entering edit mode

Thank you very much Gordon-- I now see Chapter 12 in the User's guide covers that topic.

My thought is to include the Cy3 column in the targets frame, with a dummy value, e.g. "Unused"

and the Cy5 column with "Treatment" or "Control" as above,

follow the vignette as far as normalization and design matrix creation,

and then extract a contrast specific to Cy5:

cont.matrix <- makeContrasts("Treatment-Control",levels=design)

Does this seem appropriate/reasonable?

Also, do you think the array quality weights / wtflags should be used/applied upstream in this situation?

Thanks very much again!

Dan Schnell

ADD REPLY
1
Entering edit mode

I've edited my answer. Since Cy3 is not a real channel representing actual RNA, it really should be thrown out. So you are back to a simple single channel analysis and no need for Chapter 12.

You don't necessary need wt flags at all, try it without. Simple is often best.

ADD REPLY
0
Entering edit mode

Thanks again!! I implemented the single channel approach that you outlined, w/out the wt flags, and the diagnostics & results produced all seem reasonable.

ADD REPLY

Login before adding your answer.

Traffic: 929 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6