Chip-Seq Ratio Intensity And Normalization

Question

[Snape_Ar]

I am interested in the data analysis of ChIP-Seq data. I am interested to know how to calculate the following:

1. Log2 Intensity Ratios and what does it mean ? Also, the ratio is between the replicates or between a background and a replicate and how do you calculate the intensity value for each sample?

2. I know how the local regression (lowess) works but why is smoothing needed for ChIP-Seq data. Is it needed to remove the non-specific binding peaks ?

3. What does a genome-wide mean in the ChIP-Seq sample means ?

The questions are based on the Paper that was published in Nature and link of the paper is as follows: http://www.nature.com/nature/journal/v471/n7339/full/nature09725.html.

Any hints or advice will be highly appreciated.

Thanks.

[Ryan Dale]

FYI, the in paper you refer to, they used ChIP-chip (tiling microarrays) -- but in your question you ask about ChIP-seq

Answer 1

[brentp]

1. The log-ratio is the ratio of the 2 channels in the microarray. See, e.g.: http://en.wikipedia.org/wiki/DNA_microarray#Two-channel_vs._one-channel_detection. Nearly all software packages will do this and other normalization steps for you.
2. Since the probes are often overlapping (http://en.wikipedia.org/wiki/Tiling_array), lowess is used to smooth adjacent / overlapping peaks.
3. ??

[Snape_Ar]

1. I understand the log-ratio of 2 channels in microarray but how do you calculate it with the ChIP-Seq data.
   
2. Thanks.
3. I think even though the total intensities are not very well normalized. So after calculating the intensity values,and shift them so that the mean intensity value across the genome is 0. In other words, we adjust the intensity values by subtracting the mean of intensity values across all tiled regions of the genome.

But my question still remains how to calculate the intensity values ?

Anyways thanks. I appreciate it.