hmChIP: a database and web server for exploring publicly available human and mouse ChIP-seq and ChIP-chip data

The ChIP-chip data were processed and normalized globally using TileProbe[1]. TileProbe is an algorithm that improves upon MAT[2] by modeling residual probe effects using a large number of publicly available ChIP samples. The output of TileProbe is backgroundcorrected log2 probe intensities. These intensities can measure protein-DNA binding strengths in a single sample without referring to control samples. They are also automatically normalized across all samples.
The background-corrected intensities of all probes in a sample are called the sample’s protein-DNA binding intensity profile. hmChIP stores these profiles for all ChIP-chip samples (including IP and control samples).

For each experiment, we also generated a peak list by applying CisGenome ChIP-chip peak calling procedure[3] to TileProbe background-corrected probe intensities (W=5, BandWidth=300, MaxGap=300, MinLen=100, MinProbe=5). Peaks with false discovery rate (FDR) below 10% were retained. During the peak detection procedure, a log2 fold change between ChIP and control was computed for each probe, after taking the average of intensities across biological and technical replicates. This resulted in a log2 fold change profile for each experiment, which was also stored in hmChIP. If an experiment does not have control samples, the log2 fold change is just the average of TileProbe background-corrected log2 probe intensities across ChIP samples.
In summary, each ChIP-chip experiment in hmChIP is associated with one peak list, one log2 fold change profile, and one or more samples. Each sample in the experiment is associated with a proteinDNA binding intensity profile. The peak list is stored in COD format, which is a standard file format recognized by CisGenome and can be used as input for a number of CisGenome analysis functions (e.g. motif mapping). The log2 fold change of the experiment and protein-DNA binding intensity profiles of individual samples are stored in BAR format, which can be visualized in CisGenome Browser and can be converted to TXT or WIG format using CisGenome[3].

[1]Judy, J.T. and Ji, H. (2009) TileProbe: modeling tiling array probe effects using publicly available data. Bioinformatics, 25, 2369-2375.

[2]Johnson, W.E., Li, W., Meyer, C.A. et al. (2006) Model-based analysis of tiling-arrays for ChIP-chip. Proc Natl Acad Sci USA, 103, 12547-12462.

[3]Ji, H., Jiang, H., Ma, W., et al. (2008) An integrated software system for analyzing
ChIP-chip and ChIP-seq data. Nat. Biotechnol. 26, 1293-1300.