Dinh/Dinh 2012/NOTES/2012-1-11

From ZhangLabWiki
Jump to navigation Jump to search

Analysis of 9 technical replicates from HL111 (NA12156)[edit]

  • Noi performed 9 capture reactions from the same bisulfite converted gDNA, and this library was captured and sequenced in HL111.
  • Noi mapped all the data and created the methylation matrix file from the 9 BED files requiring data from at least 7 of the 9 replicates.
    • 7 data points is good enough to determine the methylation distribution and variance of a captured site.
    • In total, there were : 412,476 sites captured by at least 7 samples.
    • However the standard deviation for ~25% of the sites were > 0.1.
  • Checking the QC values for this dataset:
 rep1 N=9175   r=0.9111
 rep2 N=13414  r=0.9292
 rep3 N=8385   r=0.9157
 rep4 N=16172  r=0.9754
 rep5 N=17041  r=0.9738
 rep6 N=9770   r=0.9273
 rep7 N=15372  r=0.9312
 rep8 N=12460  r=0.8802
 rep9 N=12389  r=0.8949
  • Why does both N and r values vary so much?
  • As the average QC drop, which sites become more variable?
  • Where these sites variable sites localized?
  • We can go back to low quality dataset and mask these variable sites.

DMR330K Resubsetting[edit]

   #padlock    Subset name (efficiency level)   #probes
   4334 Exp1_3.s1 (A)    3428
  28560 Exp1_3.s2 (B)    23849
  33285 Exp1_3.s3 (C)    28290
  26490 Exp1_3.s4 (D)    22440
  33379 Exp1_3.s5 (E)    28064
   7328 Sub1_7.s1 (A)    3544
  56748 Sub1_7.s2 (B)    30213
  52737 Sub1_7.s3 (C)    29818
  43685 Sub1_7.s4 (D)    24747
  20620 Sub1_7.s5 (E)    11771
 307166 total
  • (A,B,C,D,E) are efficiency levels determined from http://genome-tech.ucsd.edu/LabNotes/index.php/Dinh_2011/NOTES/2011-12-27
    • Since the change in efficiency from level A to B is much higher than the changes from B to C, C to D, and D to E, probes at level A must be kept separate.
    • Although the probes at level E are not very efficient, I have already removed probes that have 0 capture in multiple capture experiments. For Sub1_7, probe placement in E is based on 0 capture in only 1 capture from the re-balancing experiment. For Exp1_3, the cutoff for E is not at 0 capture, so the probes at the top of level E are still good.
  • Combine probes at levels B, C, and D (probes in E are not ranked properly since they have 0 capture in 1 experiment)
cat Sub1_7.s2.txt.seq Sub1_7.s3.txt.seq Sub1_7.s4.txt.seq > tmp
split -l 55000 tmp DMR220K.
  55000 DMR220K.aa (B)
  55000 DMR220K.ab (C+D)
  43170 DMR220K.ac (D) 
  • Go into each of the 55K files and make sure that the split is not between variants of the same probe.
  • Fill up DMR220K.ac to make 55K:
 tail -n 11830 DMR220K.ac > tmp
 cat DMR220K.ac tmp > DMR220K.ac+
  • There are 20,620 probes in level E which can't be ranked. I will add 4334 probes from Exp1_3 level A to this which can be amplified with an alternate reverse primers.
 tail -n 2046 Sub1_7.s5.txt.seq > tmp
 cat Sub1_7.s5.txt.seq tmp > DMR220K.ad
 ./probe2padlockCpGLibJan11.pl Exp1_3.s1.txt altAP2
 cat DMR220K.ad Exp1_3.s1.txt.seq > Mixed_ExpA.DmrE
  • DMR220K (Sub1_7) have been subsetted into 3x55K sets. with Mixed_ExpA.DmrE (27K) set.
  • Combine probes at levels B, C, D, and E and then split into subsets of 55K for Exp1_3:
cat Exp1_3.s2.txt.seq Exp1_3.s3.txt.seq Exp1_3.s4.txt.seq Exp1_3.s5.txt.seq > tmp
split -l 55000 tmp EXP.
   55000 EXP.aa (B + C)
   55000 EXP.ab (C + D + E)
   11714 EXP.ac (E)
  • Go into each of the 55K files and make sure that the split is not between variants of the same probe.
  • Add 7958 duplicates in EXP.ac (E) with itself to create doubles: 11714 + 7958 (repeats) + 7328 (DmrA) = 27K
  ./probe2padlockCpGLibJan11.pl Sub1_7.s1.txt altAP2
  tail -n 7958 EXP.ac > tmp
  cat EXP.ac tmp Sub1_7.s1.txt.seq > Mixed_DmrA.ExpE

Ready to order padlock probes[edit]