Dinh/Dinh 2014/NOTES/2014-4-1: Difference between revisions

From ZhangLabWiki
Jump to navigation Jump to search
>Dinh
>Dinh
 
(47 intermediate revisions by the same user not shown)
Line 1: Line 1:
=DMR finding with MOdel-based Analysis of Bisulfite Sequencing data (MOABS)=
=DMR finding with MOdel-based Analysis of Bisulfite Sequencing data (MOABS)=
* Data mapping statistics: [[File:Mouse_WGBS_DMR_mapping.xls]]
* We obtained additional ESCs data from 3 sources:
  AL.mESC (Alice's mESC, same strain as IPS and SCNT in dataset)C57BL/6J Jcl
  TO.mESC http://www.ncbi.nlm.nih.gov/biosample/765198         C57BL/6J Jcl
  ST.mESC http://www.ncbi.nlm.nih.gov/biosample/SAMN00631233    Mixed (129-C57Bl/6)
  HA.mESC http://www.ncbi.nlm.nih.gov/biosample/SAMN01797652    129/Ola
* Using only the mcomp module from the MOABS pipeline requires:
* Using only the mcomp module from the MOABS pipeline requires:
   1. setting LC_ALL variable with:
   1. setting LC_ALL variable with:
Line 22: Line 28:
   /home/dinh/softwares/moabs-v1.2.7.src.x86_64_Linux.data/bin/mcomp -r merged.ips.$c.G.bed -r merged.scnt.$c.G.bed -c moabs.ips_scnt.$c -l ips.$c scnt.$c
   /home/dinh/softwares/moabs-v1.2.7.src.x86_64_Linux.data/bin/mcomp -r merged.ips.$c.G.bed -r merged.scnt.$c.G.bed -c moabs.ips_scnt.$c -l ips.$c scnt.$c
===Results===
===Results===
*MOABS calculated DMRs using 3 methods. '''Need to look up what the three methods are..'''
*MOABS calculated DMRs using 3 methods.
*GREAT analysis on DMRs using single gene per region association within 10kb extended from midpoint of regulatory domain.
*Method #1, # DMRs
*Method #1, # DMRs
** call Dmc and Dmr according to method 1: Fisher Exact Test
** call Dmc and Dmr according to method 1: Fisher Exact Test
Line 28: Line 35:
** Dmr is defined as minC >=3, maxDist 300, p<p_fet_cut(suggest 0.05), nMethDif > nominalMethDifCutoff(0.3333)
** Dmr is defined as minC >=3, maxDist 300, p<p_fet_cut(suggest 0.05), nMethDif > nominalMethDifCutoff(0.3333)
   22,125 esc_vs_ips [[File:dmr_M1_esc_vs_ips.txt]]
   22,125 esc_vs_ips [[File:dmr_M1_esc_vs_ips.txt]]
        (Pathways common/BioCyc pathways) protein citrullination - 5 genes/15 regions: Padi1, Padi2, Padi3, Padi4, Padi6
   22,135 esc_vs_scnt [[File:dmr_M1_esc_vs_scnt.txt]]
   22,135 esc_vs_scnt [[File:dmr_M1_esc_vs_scnt.txt]]
     709 ips_vs_scnt [[File:dmr_M1_ips_vs_scnt.txt]]
        (Pathways common/BioCyc pathways) protein citrullination - 5 genes/15 regions: Padi1, Padi2, Padi3, Padi4, Padi6
     708 ips_vs_scnt [[File:dmr_M1_ips_vs_scnt.txt]]
        No pathway, but there were 11 significant terms for Mouse Phenotype:
        [[File:dmr_M1_ips_vs_scnt_mousephenotype.png | 300px]]
    
    
*Method #2, # DMRs
*Method #2, # DMRs
Line 36: Line 47:
**Dmr is defined as minC >=3, maxDist 300, cMethDif > credibleMethDifCutoff(0.2)
**Dmr is defined as minC >=3, maxDist 300, cMethDif > credibleMethDifCutoff(0.2)
   19,799 esc_vs_ips [[File:dmr_M2_esc_vs_ips.txt]]
   19,799 esc_vs_ips [[File:dmr_M2_esc_vs_ips.txt]]
        GO Molecular Function
              protein-arginine deiminase activity,
              hydrolase activity (acting on carbon-nitrogen) but not peptide bonds
              IgG binding
        (Pathways common/BioCyc pathways) protein citrullination - 5 genes/13 regions: Padi1, Padi2, Padi3, Padi4, Padi6
   15,318 esc_vs_scnt [[File:dmr_M2_esc_vs_scnt.txt]]
   15,318 esc_vs_scnt [[File:dmr_M2_esc_vs_scnt.txt]]
       53 ips_vc_scnt [[File:dmr_M2_ips_vs_scnt.txt]]
        GO Molecular function
              protein-arginine deiminase activity
        GO Biological process
              peptidyl-citrulline biosynthetic process from peptidyl-arginine
        (Pathways common/BioCyc pathways) protein citrullination - 5 genes/12 regions: Padi1, Padi2, Padi3, Padi4, Padi6
          BioCyc pathways
              phosphate acquisition II
       52 ips_vc_scnt [[File:dmr_M2_ips_vs_scnt.txt]]
          No term


*Method #3, # DMRs
*Method #3, # DMRs
**call Dmc according to method 2
**call Dmc according to hidden markov model with 3 hidden states: hyper/hypo/no_diff.
**Dmc is defined as cMethDif > credibleMethDifCutoff(0.2)
**the states are determined by the difference of true methylation ratios bigger than some value. The default value is 0.2
**call Dmr using Hidden Markov Model
   20,124 esc_vs_ips [[File:dmr_M3_esc_vs_ips.txt]]
   20,124 esc_vs_ips [[File:dmr_M3_esc_vs_ips.txt]]
        GO Molecular function
              protein-arginine deiminase activity
        (Pathways common/BioCyc pathways) protein citrullination - 5 genes/11 regions: Padi1, Padi2, Padi3, Padi4, Padi6
   15,550 esc_vs_scnt [[File:dmr_M3_esc_vs_scnt.txt]]
   15,550 esc_vs_scnt [[File:dmr_M3_esc_vs_scnt.txt]]
        GO Molecular function
              protein-arginine deiminase activity
        GO Biological process
              peptidyl-citrulline biosynthetic process from peptidyl-arginine
        (Pathways common/BioCyc pathways) protein citrullination - 5 genes/13 regions: Padi1, Padi2, Padi3, Padi4, Padi6
       21 ips_vc_scnt [[File:dmr_M3_ips_vs_scnt.txt]]
       21 ips_vc_scnt [[File:dmr_M3_ips_vs_scnt.txt]]
        GO Biological Process
              negative regulation of vascular endothelial growth factor receptor signaling pathway
              negative regulation of cellular response to growth factor stimulus
* The SCNTs appear to be more similar epigenetically to iPSCs than to ESCs.  
* The SCNTs appear to be more similar epigenetically to iPSCs than to ESCs.  
* I converted the DMR files to bed format and uploaded them to GREAT for enrichment analysis. For both sets of DMRs between esc and ips/scnt, the pathway that comes up is '''protein citrullination'''.  
* I converted the DMR files to bed format and uploaded them to GREAT for enrichment analysis. For both sets of DMRs between esc and ips/scnt, the pathway that comes up is '''protein citrullination'''.  
* I looked up the term and found an interesting excerpt from Wikipedia:
* I looked up the term and found an interesting excerpt from Wikipedia:
** Citrullination is important because it controls the expression of genes, particularly in the developing embryo, and because the immune system often attacks citrullinated proteins, leading to autoimmune diseases such as rheumatoid arthritis and multiple sclerosis.
** Citrullination is important because it controls the expression of genes, particularly in the developing embryo, and because the immune system often attacks citrullinated proteins, leading to autoimmune diseases such as rheumatoid arthritis and multiple sclerosis.
=== Citrullination: Padi4 gene methylation ===
* Genome coordinate of Padi4: chr4:140,301,423-140,330,118 (anti-sense strand)
* Methylation plot
[[File:my_methyl_Padi4_graph.pdf]]
* Labels:
  mES1_serum    B6ES
  mES2_2i        Habibi_ES_2i_adapted
  mES3_2i        Habibi_ES_2i
  mES4_serum    Habibi_ES_serum
  miPS1_serum    miPS_1E12P20
  miPS2_serum    miPS_2A4F1
  miPS3_serum    miPS_2A4F33
  miPS4_serum    miPS_B3
  mSCNT1_serum  SCNT_B12_serum_adapted
  mSCNT2_serum  SCNT_NB3_serum_adapted
  mSCNT3_2i      SCNT_ntesP7C_2i
  mSCNT4_2i      SCNT_ntesP8B_2i
* Plotting script
[[File:plot_methyl_graph_PADI4.txt]]
* By visual inspection, we saw multiple DMRs associated with this gene. SCNTs in serum condition is epigenetically different from ESC/IPS.
===False discovery rate===
===False discovery rate===
* which method have the lowest false discovery rate?
* A) samples were randomly assigned to either group 1 or group 2.
* shuffle sample labels
* B) samples were randomly assigned to either group and the positions were randomly assigned (while keeping the chromosome assignments constant.)
=== Next steps ===
* ESC vs IPS
* Determine whether method 1, 2 or 3 is more appropriate.
              A      B
** Generate plots to see whether the DMRs are also detectable by eye.
  Method #1    56.20%  2.70%
* Does the lower depth of coverage for the non-ESC samples (Habibi et al data have 30x average depth of coverage while our data have between 7-10 average depth of coverage) affect the number of DMRs detected for iPSCs versus SCNTs ? I need to downsample the Habibi and Stadler data.
  Method #2    47.90%  2.23%
* Use MOABS to find DMRs between 2i ESC and 2i SCNTs
  Method #3    45.77%  0.92%
* ESC vs SCNT
              A      B
  Method #1    88.12%  6.84%
  Method #2    75.52%  5.38%
  Method #3    72.32%  2.03%
* IPS vs SCNT
              A      B
  Method #1    89.69%  2.51%
  Method #2    52.31%  0.19%
  Method #3    45.24%  0%
* Method #3 have the lowest FDR.
=== Down-sampling of the Habibi and Stadler ESC data ===
* From visual inspection, Alice noticed that many ESC vs SCNT/IPS DMRs have coverage only in Habibi/Stadler data.
* Downsample Habibi data to 30% and Stadler data to 70%.
* ESC vs SCNT
      # DMRs  %of previous total DMR called
  M1  10720    48%
  M2    5049    33%
  M3    4904    32%
 
* ESC vs IPS
      # DMRs  %of previous total DMR called
  M1  11542    52%
  M2    7622    38%
  M3    8632    43%
* From this, we decided that since currently MOABs does not consider replicates information, we need to do individual ESCs comparisons
  A) AL.Indx11 vs SCNT/IPS
  B) TO.mESC vs SCNT/IPS
  C) ST.mESC vs SCNT/IPS
=== MOABS DMRs: individual ESCs versus SCNT/IPS, IPSCs versus SCNTs, SCNTs in 2i versus serum ===
* After the first run, there were very few DMRs since AL.Indx11 have low coverage, so I reduced the minCredibleDiff from 0.2 to 0.15
* These DMRs are generated from Method #2 (M2), the minimum depth requirement is set at default (=3), and is recommended for method 2.
** We found out that since the replicates for SCNT/IPS are merged, there minimum depth per sample is not 3, so to have greater confidence for the DMRs, we filtered each GBED file to allow only minimum 5x depth of coverage per site.
* AL.Indx11 (B6)
  ESC vs IPS    218 (3.12% FDR) [[File:AL-Indx11_vs_IPS_M2_MOABS-dmrs.txt]]
  ESC vs IPS    153 (4.64% FDR) [[File:AL-Indx11_vs_IPS_M2_MOABS-dmrs-min5depth.txt]]
  ESC vs SCNT    204 (1.13% FDR) [[File:AL-Indx11_vs_SCNT_M2_MOABS-dmrs.txt]]
  ESC vs SCNT    151 (0.99% FDR) [[File:AL-Indx11_vs_SCNT_M2_MOABS-dmrs-min5depth.txt]]
* ST.mESC (mixed strain)
  ESC vs IPS  19431 (1.59% FDR) [[File:ST-mESC_vs_IPS_M2_MOABS-dmrs.txt]]
  ESC vs IPS  16771 (1.83% FDR) [[File:ST-mESC_vs_IPS_M2_MOABS-dmrs-min5depth.txt]]
  ESC vs SCNT  14580 (3.09% FDR) [[File:ST-mESC_vs_SCNT_M2_MOABS-dmrs.txt]]
  ESC vs SCNT  11899 (2.69% FDR) [[File:ST-mESC_vs_SCNT_M2_MOABS-dmrs-min5depth.txt]]
* HA.mESC (different mouse strain)
  ESC vs IPS  43553 (4.31% FDR) [[File:HA-mESC_vs_IPS_M2_MOABS-dmrs.txt]]
  ESC vs IPS  39121 (5.52% FDR) [[File:HA-mESC_vs_IPS_M2_MOABS-dmrs-min5depth.txt]]
  ESC vs SCNT  38556 (8.83% FDR) [[File:HA-mESC_vs_SCNT_M2_MOABS-dmrs.txt]] 
  ESC vs SCNT  32806 (8.52% FDR) [[File:HA-mESC_vs_SCNT_M2_MOABS-dmrs-min5depth.txt]]
* TO.mESC (B6, This data was initially not used in previous analysis because it clusters with 2i condition, and was not cultured in serum. We're not sure if cultured in 2i)
  ESC vs IPS    5792 (0.65% FDR) [[File:TO-mESC_vs_IPS_M2_MOABS-dmrs.txt]]
  ESC vs IPS    4751 (0.50% FDR) [[File:TO-mESC_vs_IPS_M2_MOABS-dmrs-min5depth.txt]]
  ESC vs IPS    5792 (0.65% FDR) [[File:TO-mESC_vs_SCNT_M2_MOABS-dmrs.txt]]
  ESC vs IPS    5122 (0.82% FDR) [[File:TO-mESC_vs_SCNT_M2_MOABS-dmrs-min5depth.txt]]
* IPS vs SCNT (data from 4 iPSCs were merged and data from 2 SCNTs were merged)
  IPS vs SCNT    192 (0.73% FDR) [[File:IPS_vs_SCNT_M2_MOABS-dmrs.txt]]
  IPS vs SCNT    173 (0.64% FDR) [[File:IPS_vs_SCNT_M2_MOABS-dmrs-min5depth.txt]]
* SCNT 2i versus SCNT serum
** The number of DMRs is very high, is this expected because Habibi et al found striking differences between the two culture conditions as well.
** The FDR rate is high for 0.15diff so I increased the minimum to 0.20diff to lower FDR.
  2i vs Serum    91304 (10.30% FDR) [[File:SCNT_2i_vs_Serum_M2_MOABS-dmrs.txt]]
  2i vs Serum    83247 (12.03% FDR) [[File:SCNT_2i_vs_Serum_M2_MOABS-dmrs-min5depth.txt]]
  2i vs Serum    44308 (5.15% FDR) [[File:SCNT_2i_vs_Serum_M2_MOABS-dmrs-min5depth-0.20diff.txt]]
 
=== PAVIS - Genes region annotation pie charts ===
* Upload the files to PAVIS, 0 bp extension on both sides
* AL mESC vs IPS
[[File:pavis_ALmESC_vs_IPS_dmrs.png |300px]]
* AL mESC vs SCNT
[[File:pavis_ALmESC_vs_SCNT_dmrs.png|300px]]
* AL IPS vs SCNT
[[File:pavis_AL_IPS_vs_SCNT_dmrs.png|300px]]
 
=== Discussion ===
* Analysis hierarchy:
**We clustered all of the samples together (including 2i, Serum, and different strains) and found that there are more strain specific differences then method differences at the top level. We also saw 2i versus Serum differences.
**We decided that to detect method specific differences, we need to use cell lines from the same strain and same growth conditions
***(We will include the analysis using the difference strains in the supplementary information)
**Our analysis revealed method specific DMRs (at < 5% FDR) that we will go on to characterize.
*** First, we looked at the genes-region association, and determined whether the DMRs are associated with specific pathways
*** Second, we looked at the DMRs in each comparison to see whether they are in promoter, introns, exons, or intergenic
*** Third, we found X, and Y% of DMRs are hypermethylated and hypomethylated respectively, and show the heat map of those DMRs.
*** Fourth, we perform bisulfite sequencing of the samples to validate a small number of DMRs.
**We have SCNT in both 2i and Serum conditions, and found a large number of DMRs between them.
*** We characterize these DMRs, and compare our results with the results from Habibi et al. (which compared ESC in 2i versus Serum)

Latest revision as of 00:34, 1 May 2014

DMR finding with MOdel-based Analysis of Bisulfite Sequencing data (MOABS)[edit]

 AL.mESC (Alice's mESC, same strain as IPS and SCNT in dataset)C57BL/6J Jcl
 TO.mESC http://www.ncbi.nlm.nih.gov/biosample/765198	        C57BL/6J Jcl
 ST.mESC http://www.ncbi.nlm.nih.gov/biosample/SAMN00631233    Mixed (129-C57Bl/6)
 HA.mESC http://www.ncbi.nlm.nih.gov/biosample/SAMN01797652    129/Ola
  • Using only the mcomp module from the MOABS pipeline requires:
 1. setting LC_ALL variable with:
    export LC_ALL=C
 2. installing RInside for R. Open R and type:
    install.packages("RInside")
  • The input for mcomp is a G.bed file (ie), more detailed explanation see the MOABS documentations.
 #chrom	start	end	ratio	totalC	methC	strand	next	Plus	tcP	mcP	Minus	tcM	mcM
 chr5	3009585	3009587	1	1	1	-	G	+	0	0	-	1	1
 chr5	3011471	3011473	0.711	45	32	B	G	+	28	18	-	17	14

Analysis of mouse WGBS[edit]

  • Prepare the G.bed files with methylFreq2Gbed.pl (genome-miner)
 /home/dinh/scripts/methylFreq2GBED.pl 1 < [methylFreq] > [BED name]
  • We perform 3 tests:
 esc versus ipsc
 esc versus scnt
 ipsc versus scnt
  • MOABS commands:
 export LC_ALL=C
 /home/dinh/softwares/moabs-v1.2.7.src.x86_64_Linux.data/bin/mcomp -r merged.esc.$c.G.bed -r merged.scnt.$c.G.bed -c moabs.esc_scnt.$c -l esc.$c scnt.$c
 /home/dinh/softwares/moabs-v1.2.7.src.x86_64_Linux.data/bin/mcomp -r merged.esc.$c.G.bed -r merged.ips.$c.G.bed -c moabs.esc_ips.$c -l esc.$c ips.$c
 /home/dinh/softwares/moabs-v1.2.7.src.x86_64_Linux.data/bin/mcomp -r merged.ips.$c.G.bed -r merged.scnt.$c.G.bed -c moabs.ips_scnt.$c -l ips.$c scnt.$c

Results[edit]

  • MOABS calculated DMRs using 3 methods.
  • GREAT analysis on DMRs using single gene per region association within 10kb extended from midpoint of regulatory domain.
  • Method #1, # DMRs
    • call Dmc and Dmr according to method 1: Fisher Exact Test
    • Dmc is defined as depth >= minDepth(suggest 10), p<p_fet_cut(suggest 0.05), nMethDif > nominalMethDifCutoff(0.3333)
    • Dmr is defined as minC >=3, maxDist 300, p<p_fet_cut(suggest 0.05), nMethDif > nominalMethDifCutoff(0.3333)
 22,125 esc_vs_ips File:Dmr M1 esc vs ips.txt
        (Pathways common/BioCyc pathways) protein citrullination - 5 genes/15 regions: Padi1, Padi2, Padi3, Padi4, Padi6 
 22,135 esc_vs_scnt File:Dmr M1 esc vs scnt.txt
        (Pathways common/BioCyc pathways) protein citrullination - 5 genes/15 regions: Padi1, Padi2, Padi3, Padi4, Padi6 
    708 ips_vs_scnt File:Dmr M1 ips vs scnt.txt
        No pathway, but there were 11 significant terms for Mouse Phenotype:
        File:Dmr M1 ips vs scnt mousephenotype.png
 
  • Method #2, # DMRs
    • call Dmc and Dmr according to method 2: Using the credible methylation difference metric (beta-binomial) distribution.
    • Dmc is defined as cMethDif > credibleMethDifCutoff(0.2)
    • Dmr is defined as minC >=3, maxDist 300, cMethDif > credibleMethDifCutoff(0.2)
 19,799 esc_vs_ips File:Dmr M2 esc vs ips.txt
        GO Molecular Function
             protein-arginine deiminase activity, 
             hydrolase activity (acting on carbon-nitrogen) but not peptide bonds
             IgG binding
        (Pathways common/BioCyc pathways) protein citrullination - 5 genes/13 regions: Padi1, Padi2, Padi3, Padi4, Padi6
 15,318 esc_vs_scnt File:Dmr M2 esc vs scnt.txt
        GO Molecular function
             protein-arginine deiminase activity
        GO Biological process
             peptidyl-citrulline biosynthetic process from peptidyl-arginine
        (Pathways common/BioCyc pathways) protein citrullination - 5 genes/12 regions: Padi1, Padi2, Padi3, Padi4, Padi6
         BioCyc pathways 
             phosphate acquisition II
     52 ips_vc_scnt File:Dmr M2 ips vs scnt.txt
         No term
  • Method #3, # DMRs
    • call Dmc according to hidden markov model with 3 hidden states: hyper/hypo/no_diff.
    • the states are determined by the difference of true methylation ratios bigger than some value. The default value is 0.2
 20,124 esc_vs_ips File:Dmr M3 esc vs ips.txt
        GO Molecular function
              protein-arginine deiminase activity
        (Pathways common/BioCyc pathways) protein citrullination - 5 genes/11 regions: Padi1, Padi2, Padi3, Padi4, Padi6
 15,550 esc_vs_scnt File:Dmr M3 esc vs scnt.txt
        GO Molecular function
             protein-arginine deiminase activity
        GO Biological process
             peptidyl-citrulline biosynthetic process from peptidyl-arginine
        (Pathways common/BioCyc pathways) protein citrullination - 5 genes/13 regions: Padi1, Padi2, Padi3, Padi4, Padi6
     21 ips_vc_scnt File:Dmr M3 ips vs scnt.txt
        GO Biological Process
             negative regulation of vascular endothelial growth factor receptor signaling pathway
             negative regulation of cellular response to growth factor stimulus
  • The SCNTs appear to be more similar epigenetically to iPSCs than to ESCs.
  • I converted the DMR files to bed format and uploaded them to GREAT for enrichment analysis. For both sets of DMRs between esc and ips/scnt, the pathway that comes up is protein citrullination.
  • I looked up the term and found an interesting excerpt from Wikipedia:
    • Citrullination is important because it controls the expression of genes, particularly in the developing embryo, and because the immune system often attacks citrullinated proteins, leading to autoimmune diseases such as rheumatoid arthritis and multiple sclerosis.

Citrullination: Padi4 gene methylation[edit]

  • Genome coordinate of Padi4: chr4:140,301,423-140,330,118 (anti-sense strand)
  • Methylation plot
File:My methyl Padi4 graph.pdf
  • Labels:
 mES1_serum     B6ES
 mES2_2i        Habibi_ES_2i_adapted
 mES3_2i        Habibi_ES_2i
 mES4_serum     Habibi_ES_serum
 miPS1_serum    miPS_1E12P20
 miPS2_serum    miPS_2A4F1
 miPS3_serum    miPS_2A4F33
 miPS4_serum    miPS_B3
 mSCNT1_serum   SCNT_B12_serum_adapted
 mSCNT2_serum   SCNT_NB3_serum_adapted
 mSCNT3_2i      SCNT_ntesP7C_2i
 mSCNT4_2i      SCNT_ntesP8B_2i
  • Plotting script
File:Plot methyl graph PADI4.txt
  • By visual inspection, we saw multiple DMRs associated with this gene. SCNTs in serum condition is epigenetically different from ESC/IPS.

False discovery rate[edit]

  • A) samples were randomly assigned to either group 1 or group 2.
  • B) samples were randomly assigned to either group and the positions were randomly assigned (while keeping the chromosome assignments constant.)
  • ESC vs IPS
              A       B
 Method #1    56.20%  2.70%
 Method #2    47.90%  2.23%
 Method #3    45.77%  0.92%
  • ESC vs SCNT
              A       B
 Method #1    88.12%  6.84%
 Method #2    75.52%  5.38%
 Method #3    72.32%  2.03%
  • IPS vs SCNT
              A       B
 Method #1    89.69%  2.51%
 Method #2    52.31%  0.19%
 Method #3    45.24%  0%
  • Method #3 have the lowest FDR.

Down-sampling of the Habibi and Stadler ESC data[edit]

  • From visual inspection, Alice noticed that many ESC vs SCNT/IPS DMRs have coverage only in Habibi/Stadler data.
  • Downsample Habibi data to 30% and Stadler data to 70%.
  • ESC vs SCNT
      # DMRs   %of previous total DMR called
 M1   10720    48%
 M2    5049    33%
 M3    4904    32%
  • ESC vs IPS
      # DMRs   %of previous total DMR called
 M1   11542    52%
 M2    7622    38%
 M3    8632    43%
  • From this, we decided that since currently MOABs does not consider replicates information, we need to do individual ESCs comparisons
 A) AL.Indx11 vs SCNT/IPS
 B) TO.mESC vs SCNT/IPS
 C) ST.mESC vs SCNT/IPS

MOABS DMRs: individual ESCs versus SCNT/IPS, IPSCs versus SCNTs, SCNTs in 2i versus serum[edit]

  • After the first run, there were very few DMRs since AL.Indx11 have low coverage, so I reduced the minCredibleDiff from 0.2 to 0.15
  • These DMRs are generated from Method #2 (M2), the minimum depth requirement is set at default (=3), and is recommended for method 2.
    • We found out that since the replicates for SCNT/IPS are merged, there minimum depth per sample is not 3, so to have greater confidence for the DMRs, we filtered each GBED file to allow only minimum 5x depth of coverage per site.
  • AL.Indx11 (B6)
 ESC vs IPS     218 (3.12% FDR) File:AL-Indx11 vs IPS M2 MOABS-dmrs.txt
 ESC vs IPS     153 (4.64% FDR) File:AL-Indx11 vs IPS M2 MOABS-dmrs-min5depth.txt
 ESC vs SCNT    204 (1.13% FDR) File:AL-Indx11 vs SCNT M2 MOABS-dmrs.txt
 ESC vs SCNT    151 (0.99% FDR) File:AL-Indx11 vs SCNT M2 MOABS-dmrs-min5depth.txt
  • ST.mESC (mixed strain)
 ESC vs IPS   19431 (1.59% FDR) File:ST-mESC vs IPS M2 MOABS-dmrs.txt
 ESC vs IPS   16771 (1.83% FDR) File:ST-mESC vs IPS M2 MOABS-dmrs-min5depth.txt
 ESC vs SCNT  14580 (3.09% FDR) File:ST-mESC vs SCNT M2 MOABS-dmrs.txt
 ESC vs SCNT  11899 (2.69% FDR) File:ST-mESC vs SCNT M2 MOABS-dmrs-min5depth.txt
  • HA.mESC (different mouse strain)
 ESC vs IPS   43553 (4.31% FDR) File:HA-mESC vs IPS M2 MOABS-dmrs.txt
 ESC vs IPS   39121 (5.52% FDR) File:HA-mESC vs IPS M2 MOABS-dmrs-min5depth.txt
 ESC vs SCNT  38556 (8.83% FDR) File:HA-mESC vs SCNT M2 MOABS-dmrs.txt  
 ESC vs SCNT  32806 (8.52% FDR) File:HA-mESC vs SCNT M2 MOABS-dmrs-min5depth.txt
  • TO.mESC (B6, This data was initially not used in previous analysis because it clusters with 2i condition, and was not cultured in serum. We're not sure if cultured in 2i)
 ESC vs IPS    5792 (0.65% FDR) File:TO-mESC vs IPS M2 MOABS-dmrs.txt
 ESC vs IPS    4751 (0.50% FDR) File:TO-mESC vs IPS M2 MOABS-dmrs-min5depth.txt
 ESC vs IPS    5792 (0.65% FDR) File:TO-mESC vs SCNT M2 MOABS-dmrs.txt
 ESC vs IPS    5122 (0.82% FDR) File:TO-mESC vs SCNT M2 MOABS-dmrs-min5depth.txt
  • IPS vs SCNT (data from 4 iPSCs were merged and data from 2 SCNTs were merged)
 IPS vs SCNT    192 (0.73% FDR) File:IPS vs SCNT M2 MOABS-dmrs.txt
 IPS vs SCNT    173 (0.64% FDR) File:IPS vs SCNT M2 MOABS-dmrs-min5depth.txt
  • SCNT 2i versus SCNT serum
    • The number of DMRs is very high, is this expected because Habibi et al found striking differences between the two culture conditions as well.
    • The FDR rate is high for 0.15diff so I increased the minimum to 0.20diff to lower FDR.
 2i vs Serum    91304 (10.30% FDR) File:SCNT 2i vs Serum M2 MOABS-dmrs.txt
 2i vs Serum    83247 (12.03% FDR) File:SCNT 2i vs Serum M2 MOABS-dmrs-min5depth.txt
 2i vs Serum    44308 (5.15% FDR) File:SCNT 2i vs Serum M2 MOABS-dmrs-min5depth-0.20diff.txt

PAVIS - Genes region annotation pie charts[edit]

  • Upload the files to PAVIS, 0 bp extension on both sides
  • AL mESC vs IPS

File:Pavis ALmESC vs IPS dmrs.png

  • AL mESC vs SCNT

File:Pavis ALmESC vs SCNT dmrs.png

  • AL IPS vs SCNT

File:Pavis AL IPS vs SCNT dmrs.png

Discussion[edit]

  • Analysis hierarchy:
    • We clustered all of the samples together (including 2i, Serum, and different strains) and found that there are more strain specific differences then method differences at the top level. We also saw 2i versus Serum differences.
    • We decided that to detect method specific differences, we need to use cell lines from the same strain and same growth conditions
      • (We will include the analysis using the difference strains in the supplementary information)
    • Our analysis revealed method specific DMRs (at < 5% FDR) that we will go on to characterize.
      • First, we looked at the genes-region association, and determined whether the DMRs are associated with specific pathways
      • Second, we looked at the DMRs in each comparison to see whether they are in promoter, introns, exons, or intergenic
      • Third, we found X, and Y% of DMRs are hypermethylated and hypomethylated respectively, and show the heat map of those DMRs.
      • Fourth, we perform bisulfite sequencing of the samples to validate a small number of DMRs.
    • We have SCNT in both 2i and Serum conditions, and found a large number of DMRs between them.
      • We characterize these DMRs, and compare our results with the results from Habibi et al. (which compared ESC in 2i versus Serum)