Dinh/Dinh 2012/NOTES/2012-2-2
Jump to navigation
Jump to search
Paper: A strand specific high resolution normalization method for chip-sequencing data employing multiple experimental control measurements[edit]
- Allows for normalization using multiple controls: Input and IgG
- Whereas most peak callers normalize the results from peak calling, this paper proposes normalizing the data before peak calling. The algorithm uses regression modeling to normalize the data and then creating dummy reads in the "normalized" data output (this can be input to MACS directly)
- When testing MAX and FOXA1 chIP data: the percentage of peaks called using normalized data as input with the expected motifs is higher. Moreover, these regions with expected motifs would be missed when data is not normalized.
- For MAX, both IgG and Input were available as controls, and they've found that with normalization, IgG and Input behave similarly as controls.
- Installing the software:
Go to github.com search "strand specific" - the program came up as the first software download before installing with "make": sudo apt-get install libgsl0-dev in Makefile: change "opt/local" to "usr" add in : "-Wl,--no-as-needed" to get: $(CC) -Wl,--no-as-needed $(CPPFLAGS) $(DFLAGS) $(GSLFLAGS_C) $(GSLFLAGS_L) src/ftumch2.cpp $(BIN_DIR)/utils.o -o $(BIN_DIR)/ftumch2 install with make copy ftumch2 to /usr/local/bin
24 chIP data[edit]
- After mapping with BWA, all the "proper pairs" were kept.
- I used this script to convert SAM to BED and instead of keeping each paired reads separate, I combined them into 1 read spanning the physical distance between the reads: File:PEsam2BEDphysical.txt
- Sort the reads by chromosome and position:
for f in Indx*.BED do sort -k1,1 -k2,2n $f > sorted.$f done;
- Normalize with ftumch2
- Use MACS to call peak
- Compare the peak called in different datasets (called peaks should span into repeats!)
- How many peaks are in repeats
- Are there differential enrichment?