Noi/NOTES/2012-3-23

From ZhangLabWiki
Jump to navigation Jump to search

Link to calendar: [[1]]

  • Continued from 2012_03_21: [[2]]
  • Since there are different numbers of SNP-free ASM and SNP-containing CpG ASM between Binomial and T-test (on the opposite pattern as shown in table on 2012_03_21). I have discussed with Dinh, and she realized that there was some issues about the criteria she used in asmBinomialTest.pl script. Dinh make the correction to the script and I repeated analysis for binomial test again.

Repeated binomial test

./asmBinomialTest.pl UCLA_MFASM_Qcpg_list.txt > UCLA_SZ_SeqDepTest_Binomial_PVal0.05.txt
  • I got 45,765 significant ASM site, p-value 0.05 and 41,962 site, p-value 0.01.
  • Then I checked the overlapping between the sites from T-test and binomial test. There are 8,889 sites overlap between the two tests which accounts for 94.3% of the sites from the T-test.
Repeated binomial test p-value 0.01 % of ASM p-value 0.05 % of ASM
SNP-free ASM 28,398 67.67% 32,051 70.03%
SNP on CpG 13565 32.33% 13,714 29.97%
Total 41,963 45,765
  • Note: # of sites have been test:
    • Binomial test: 94,093 sites
    • TTest: 73,271 sites
./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

Discussion with Dr. Zhang[edit]

  • After discussing with Dr. Zhang, he seems to like the idea of using T-test to characterize a significant ASM rather than binomial test since the significant sites tend to show a consistency of sequence preference DNA methylation.
  • He also suggested to calculate FDR with the two options (FDR < 20% is still acceptable, but <5% is ideal)
    • Using MULTTest (R)
    • Calculate FDR = (N * p-value cutoff)/M, N = number of CpG sites have been tested, M = number of significant CpG site derived from the test at p-value cutoff
      • I have calculate FDR,
p-value cutoff 0.05 -->(73,271 * 0.05)/9,123 = 40.1% FDR (this number seem to be very high FDR)
p-value cutoff 0.01 -->(73,271 * 0.01)/8,281 = 8.8% FDR
p-value cutoff 0.02 -->(73,271 * 0.02)/8,563 = 17.1% FDR
      • I will try to calculate FDR by MULTTest function in R
  • Another suggestion from him is to plot the graph between -log(p-value) vs. the distance between SNP and CpG (for those site of SNP-free CpG), and see the trend of the graph.
  • More details for suggestions for mpo-mQTL-ASM analysis: Dr. Zhang suggested to generated Ven diagram to see the overlapping of CpG sites from three analyses.
  • Another issue is make sure that the SNP and CpG coordinates are consistent since this could affect the number of SNP-free and SNP-containing CpG on ASM analysis.
    • 2012_03_26 Check the coordinates of SNP and CpG showing ASM on genome browser. Ex. Chr5:43194137(SNP):43194136(CpG) for CpG position if SNP and CpG were assigned on the same coordinate (both start position (0) or both stop position (1), CpG should show C while SNP should show any sequence A G T or C, but not always the same sequence as CpG. Otherwise it is likely that CpG is assigned as1 and SNP is assigned as 0 coordinate.
    • After checking multiples SNP-CpG pairs, I am pretty sure that their coordinates were consistent. Ex. Chr5:43194137(SNP) --> G:43194136(CpG) --> C, Chr4:5990551(SNP)--> G:5990550(CpG) --> C, Chr9:139000412(SNP) --> C :139000411 --> C, Chr1:26485013(SNP) --> T :26485012(CpG) --> C
    • Also if SNP is on CpG it should show the same sequence. Ex. Chr15:100104385(SNP) --> C :100104385(CpG) --> C, Chr11:17461329(SNP) --> C :17461329(CpG) --> C

mpo and mQTL analysis (have been repeated on 2012_03_26)[edit]

  • I have selected the CpG sites from mpo analysis and mQTL analysis that overlap to see if genetic effect contributes to DNA methylation status.
  • I got 104 of CpG sites that overlap and performed pathway enrichment analysis with GREAT and got a list of genes and the distance to TSS of those sites to the genes.

Media:UCLA_SZ_heriatble_CpG-SNP_assoc_noSNP-parsegeneList-min2sites.txt
Media:UCLA_SZ_heriatble_CpG-SNP_assoc_noSNP-parsegeneList.txt