Athurva Gore/LabNotes/2009-8-27: Difference between revisions

From ZhangLabWiki
Jump to navigation Jump to search
>Ajgore
(New page: {{ AGLabEntry|2009-8-26|2009-8-28 }} =To Do= * Forms for Tech Transfer Office * Invoice for Zymo =Library-free Probesets= * Investigate TDMR crash today; can we just skip that one site? ...)
 
>Ajgore
No edit summary
 
(7 intermediate revisions by the same user not shown)
Line 14: Line 14:
** Reducing quality scores of end bases when SNPs found
** Reducing quality scores of end bases when SNPs found
** Reducing quality scores of all end bases
** Reducing quality scores of all end bases
==NA12878==
* Looked at HL006_s5 data
** Wanted to compare the small number of misses remaining to 1000 genome.
* Found that if a coverage filter of >4 is set,
hits misses miscalls
Total Number
29      111        7
Number with coverage < 4:
4      86        7
* Lose only four hits, but eliminate most of the misses and all of the miscalls!
** This makes sense, as with coverage < 4 (2 or 3 reads), we cannot be certain of the site.
==DF-6-9-9==
* Hoping to get a filter that allows good coverage but has ~5% of SNPs outside dbSNP only
* Can we accomplish this simply by thresholding coverage?
* Looked at DF699 from HL004 with a 5-bp tossout range.
** Has more hits; want to see if we can filter misses at this stage.
** If we filter by coverage < 4:
** Lose 2108/11314 hits, 666/2139 misses
* We do better if we cut out more reads (with a 15-bp tossout range)
** Lose 1944/11314 hits (9370), 1040/2139 misses (1099)
* Investigating the reads is probably a better strategy, though post-processing out low-coverage SNPs is probably not a bad idea.
=Trimming bad reads=
* Wrote a script that behaves in the same way as the rejectBadLigation.pl, but instead trims reads down to the correct size
* Takes a variable input as well
* With DF699 from HL004, reduces error.
** 10061 hits, 1376 misses
** Not as much reduction as removing reads; more hits kept, however

Latest revision as of 20:17, 28 August 2009

Navigation[edit]

To Do[edit]

  • Forms for Tech Transfer Office
  • Invoice for Zymo

Library-free Probesets[edit]

  • Investigate TDMR crash today; can we just skip that one site?

IPS and Cancer[edit]

  • Want to test various trimming methods to see which works best.
    • Removing end bases from reads with SNPs
    • Removing end bases from all reads
    • Reducing quality scores of end bases when SNPs found
    • Reducing quality scores of all end bases

NA12878[edit]

  • Looked at HL006_s5 data
    • Wanted to compare the small number of misses remaining to 1000 genome.
  • Found that if a coverage filter of >4 is set,
hits misses miscalls
Total Number		
29      111        7
Number with coverage < 4:		
4       86         7
  • Lose only four hits, but eliminate most of the misses and all of the miscalls!
    • This makes sense, as with coverage < 4 (2 or 3 reads), we cannot be certain of the site.

DF-6-9-9[edit]

  • Hoping to get a filter that allows good coverage but has ~5% of SNPs outside dbSNP only
  • Can we accomplish this simply by thresholding coverage?
  • Looked at DF699 from HL004 with a 5-bp tossout range.
    • Has more hits; want to see if we can filter misses at this stage.
    • If we filter by coverage < 4:
    • Lose 2108/11314 hits, 666/2139 misses
  • We do better if we cut out more reads (with a 15-bp tossout range)
    • Lose 1944/11314 hits (9370), 1040/2139 misses (1099)
  • Investigating the reads is probably a better strategy, though post-processing out low-coverage SNPs is probably not a bad idea.

Trimming bad reads[edit]

  • Wrote a script that behaves in the same way as the rejectBadLigation.pl, but instead trims reads down to the correct size
  • Takes a variable input as well
  • With DF699 from HL004, reduces error.
    • 10061 hits, 1376 misses
    • Not as much reduction as removing reads; more hits kept, however