Athurva Gore/LabNotes/2009-8-24

From ZhangLabWiki
Jump to navigation Jump to search

Navigation[edit]

Probe Sets[edit]

  • FLYDUP5279 has finished; all old probesets are done
  • Need to create and send invoice to Zymo today
    • Also mention that probe generation can be rerun without quality filters?
  • Need to redo TDMR set and Cpg-SNP set, but with a new probe design...
    • Constrain arm+target size to 105 bp
    • Then modify P2MIP_v9 to use a 45 bp total arm size
    • This will allow library construction to not be necessary
      • During capture, MME1 digestion removes 17-18 bp from each side of H1_TARGET_H2
      • Should ideally leave us with 75 bp segment remaining, 60 bp of which will be target
  • This can be read in the sequencer without library construction, allowing for a much quicker and more efficient experimental procedure

Exome Read Analysis[edit]

  • Turns out that HL_003 had very poor focus quality. Cannot trust the data because of this, as calls may have been mislocated.
  • Flowcell was reused for HL_006; same capturing experiment, but much higher quality data (microscope was in focus)
  • Analyzing reads from HL_006
    • Comparing Lane 4 and Lane 5
    • Lane 4 - NA12878, size-selected
    • Lane 5 - NA12878, no size selection
  • Mapped using Bowtie, called SNPs using SAM
  • Compared to 1KG and dbSNP
  • Results BEFORE ANY FILTERING:
    • Lane 4 - Size Selection
      • vs 1KG, 4,044 hits, 1,533 misses, 218 miscalls (right location but wrong call)
      • vs dbSNP, 4,309 SNPs in dbSNP, 1,492 not in dbSNP
    • Lane 5 - No Size Selection
      • vs 1KG, 3,935 hits, 21,493 misses, 305 miscalls (right location but wrong call)
      • vs dbSNP, 4,411 SNPs in dbSNP, 21,328 not in dbSNP
  • Wanted to investigate why this was happening.
  • Could junk be getting appended to reads, causing successful mapping to produce errors?
    • HERE, TESTED ONLY AGAINST DBSNP.
    • Hypothesis: Extra bases are being ligated onto the start or end of captured regions, causing mapping to be successful but produce many homozygous SNPs
    • Checked both s4 (size selected) and s5 (not size selected)
    • Results, no filtering
      • S4: Misses have on average 28% start/end reads, Hits have on average 7% start/end reads
      • S5: Misses have on average 42% start/end reads, Hits have on average 8% start/end reads
    • Bases being added to beginning or end of reads could be causing issues?
    • Checked the reads themselves.
    • Saw a large number of reads with an error in the "39th" position (ie either the start or end of mapped read depending on orientation)
      • S5 (not size-selected) had 153286, while S4 (size-selected) had 50140
      • Size-selected version had much fewer of these errors.
    • Could it just be 3' sequencing problems?
      • Would not make as much sense, as multiple reads had the exact same error for many of the erroneous SNP calls.
      • Would get results like 30-40% reads are at start/end position? Could be artifacts from low-coverage SNPs.
    • Checked with a coverage filter of > 4:
      • S4: Misses have on average 12% start/end reads, Hits have on average 7% start/end reads
      • S5: Misses have on average 29% start/end reads, Hits have on average 6% start/end reads
    • This is a bit harder to read. It appears that problem is no longer existent in size-selected case, but with no size-selection, problem still exists.
  • CHECK WHILE REMOVING CLONAL READS
    • This statistic is most likely due to clonal reads still being counted under Dr. Zhang's script.
    • Tried re-running on s4 and s5 using SAMtools' rmdup to remove clonal reads
    • Check if this works.
  • rmdup does not seem to be working
    • Modified downSampler code to only keep one of each read type
    • Then check calls; see if we do better
    • Might do a lot better, as many errors seem to arise from ends of reads...
  • No longer have multiple reads starting and ending in the same position.
    • Should hopefully perform better
    • Now comparing to 1KG to check