Noi/NOTES/2012-3-21
Jump to navigation
Jump to search
Link to calendar: [[1]]
ASM analysis of UCLA SZ data set
- After discussing with Dr. Zhang and Dinh about how to identify sequence-dependent ASM, Dr. Zhang suggested to identify the consistency of sequence preference DNA methylation across individual
- After having more discussion with Dinh, she helped me out to write scripts for sequence-dependent ASM using binomial test and t-test by testing on ASM HAPMAP data (total 12 samples): [[2]]
on genome-miner: /home/nplongth/Noi_scratch/ASM_HAPMAP1362-1454_2012_02_03/DD_SequenceDependentTest
- I did the same analysis on UCLA SZ data set
on genome-miner: /home/nplongth/Noi_scratch/ASM_UCLA/combined_ASM_UCLA-4batches_2012_01_24
- Since there were mislabeling of some samples at the beginning from UCLA and after mapping and ASM analysis, and I generated many files. I still kept all labeling the same as original, but made the correction when generated the methylation matrix or other sample_list files for any analysis. Here are the list of samples used in analysis and correction.
Correct data mislabeling in UCLA.SZ data set
Note: --> = change the labeling to
GK0210-001 --> GK0210-002
GK0210-002 --> GK0210-001
GK0050-005a --> GK0277-002
GK0277-002 --> GK0344-004
GK0344-004 --> GK0050-005a
- Generated the list of MFASMAnalysisQValues.GK0XXXXXX.cpg.txt in UCLA_MFASM_Qcpg_list.txt: Media:UCLA_MFASM_Qcpg_list.txt
- Note: For these two tests (binomial and t-test), we require minimum 5 data points and p-value of 0.05
Binomial test
- Performed binomial test using asmBinomialTest.pl script written by Dinh: Media:asmBinomialTest.txt
./asmBinomialTest.pl UCLA_MFASM_Qcpg_list.txt > UCLA_SZ_SeqDepTest_Binomial_PVal0.05.txt
- I got total 8,510 SNP:CpG sites showing significant ASM using p-value cutoff 0.05. However, when I looked closer to the methylation level on each allele showing significant ASM by this test, it showed that some sites contain very close methylation level or have the methylation fraction 0 on both alleles. I may try to increasing stringency by lowering the p-value and observe the resulting methylation level on each allele.
awk '{if ($98 <=0.01) print $0;}' UCLA_SZ_SeqDepTest_Binomial_PVal0.05.txt > UCLA_SZ_SeqDepTest_Binomial_PVal0.01.txt
- After lowering p-value cutoff, I got total 4,708 SNP:CpG sites.
- Note: In the first column of the table, 13:38445196:38445195 --> chromosome position: SNP position: CpG position
T-test
- Performed t-test using asmTTest.pl script written by Dinh: Media:asmTTest.txt
./asmTTest.pl UCLA_MFASM_Qcpg_list.txt > UCLA_SZ_SeqDepTest_TTest_PVal0.05.txt
- I got total 9,123 SNP:CpG sites showing significant ASM, p-value 0.05.
awk '{if ($98 <=0.01) print $0;}' UCLA_SZ_SeqDepTest_TTest_PVal0.05.txt > UCLA_SZ_SeqDepTest_TTest_PVal0.01.txt
- I got 8,281 SNP:CpG sites, p-value 0.01
- --> explain more
- Extract number of significant ASM
- SNP-free ASM:
- SNP on CpG ASM:
./getnoSNPCpG.pl UCLA_SZ_SeqDepTest_Binomial_PVal0.01.txt > UCLA_SZ_SeqDep_Binomial-noSNP_PVal0.01.txt ./getSNPonCpG.pl UCLA_SZ_SeqDepTest_Binomial_PVal0.01.txt > UCLA_SZ_SeqDep_Binomial-SNPonCpG_PVal0.01.txt
./getnoSNPCpG.pl UCLA_SZ_SeqDepTest_Binomial_PVal0.05.txt > UCLA_SZ_SeqDep_Binomial-noSNP_PVal0.05.txt ./getSNPonCpG.pl UCLA_SZ_SeqDepTest_Binomial_PVal0.05.txt > UCLA_SZ_SeqDep_Binomial-SNPonCpG_PVal0.05.txt
./getnoSNPCpG.pl UCLA_SZ_SeqDepTest_TTest_PVal0.01.txt > UCLA_SZ_SeqDep_TTest-noSNP_PVal0.01.txt ./getSNPonCpG.pl UCLA_SZ_SeqDepTest_TTest_PVal0.01.txt > UCLA_SZ_SeqDep_TTest-SNPonCpG_PVal0.01.txt
./getnoSNPCpG.pl UCLA_SZ_SeqDepTest_TTest_PVal0.05.txt > UCLA_SZ_SeqDep_TTest-noSNP_PVal0.05.txt ./getSNPonCpG.pl UCLA_SZ_SeqDepTest_TTest_PVal0.05.txt > UCLA_SZ_SeqDep_TTest-SNPonCpG_PVal0.05.txt
- Comparing the the results from Binomial test and T-test
' | Binomial test | T-test | overlapped Binomial/T-test |
p-value 0.01 | |||
SNP-free ASM | 3,898 | 639 | 38 |
SNP on CpG ASM | 810 | 7,642 | 366 |
Total | 4,708 | 8,281 | 404 |
p-value 0.05 | |||
SNP-free ASM | 7,502 | 978 | 156 |
SNP on CpG ASM | 1,008 | 8,145 | 466 |
Total | 8,510 | 9,123 | 622 |
- Form these numbers it suggested that there were less than 10% of the significant ASM were consistent by the two tests.
./getnoSNPCpG.pl overlapped_UCLA_SZ_SeqDep_BinoTTest_PVal0.01.txt > overlapped_UCLA_SZ_SeqDep_BinoTTest-noSNP_PVal0.01.txt ./getSNPonCpG.pl overlapped_UCLA_SZ_SeqDep_BinoTTest_PVal0.01.txt > overlapped_UCLA_SZ_SeqDep_BinoTTest-SNPonCpG_PVal0.01.txt
./getnoSNPCpG.pl overlapped_UCLA_SZ_SeqDep_BinoTTest_PVal0.05.txt > overlapped_UCLA_SZ_SeqDep_BinoTTest-noSNP_PVal0.05.txt ./getSNPonCpG.pl overlapped_UCLA_SZ_SeqDep_BinoTTest_PVal0.05.txt > overlapped_UCLA_SZ_SeqDep_BinoTTest-SNPonCpG_PVal0.05.txt
- --> describe more
T-Test: plotting significant ASM
- SNP-free ASM examples:
File:UCLA ASM TTest-1.png File:UCLA ASM TTest-2.png p-value: Number of data point
- SNP on CpG ASM example: