Detection P-value in illumina data

Question

[parinv]

hello,

I have started to work with an Illumina dataset [platform:-GPL10558-Illumina HumanHT-12 V4.0 expression beadchip]. I don't have any prior knowledge of Illumina data analysis. I imported my file to R-studio using lumiR function of the Lumi package but the data has some sample replicate and Detection P-value.

Can anyone suggest how I remove these replicate samples and analyze the Illumina data with P-values in it? Do I need to remove the detection P-value or use them in the analysis?

Thank you

Parin

[DareDevil]

Detection p value is a concept (mainly) used by Illumina. There is a set of negative controls on Illumina BeadArrays that are used to compute the background intensity. A small detection p value indicates that the measured intensity is very likely to be a true (significant) signal and not background noise. The intensity associated with a certain detection p value corresponds to the respective upper quantile (percentile) of the (probability) distribution of the intensities of the negative controls.

Answer 1

[Kevin Blighe]

Based on voting patterns, it seems that you found some other posts on this topic. Is this correct?

[parinv]

yes, I found some questions related to illumina data but that did not clear my confusion regarding the detection P value.

[Kevin Blighe]

You can use the detection p-value to filter out probes that failed. This is reported in the limma manual, I think.

[parinv]

After filtering the probes, should we remove the detection p-value columns or keep them for further analysis?

[Kevin Blighe]

You will not need them for anything that is downstream of the probe filtering

[parinv]

Thankyou Kevin, I just have one last question. I Came across some papers, in which they filtered probes before background correction. Can you please explain if it is the correct way to filter probes and why should we background correct the data and then filter probes?

[Kevin Blighe]

Please elaborate on which specific filtering was performed in the papers that you mention.

The detection p-values are used during normalisation to determine precisely those probes that provide reliable signal. Please see the entire pipeline (advanced) which I posted recently: A: GPL6883_HumanRef-8_V3_0_R0_11282963_A (illumina expression beadchip)

Background correction and normalisation is performed via the neqc() function, which takes into account the detection p-values if the input object is setup correctly - it will issue a message to terminal stating just this.

[parinv]

thank you so much Kevin for your help.