Ylaine/2009-8-18

From ZhangLabWiki
Jump to navigation Jump to search

Finding Targeted SNPs[edit]

  • New script, snpTargeted.pl adapted from snpInRange.pl
../Scripts/snpTargeted.pl orig.snp>snp.targeted&
  • Got 17557/20761 or 84.57%
  • Follow-up questions:
    • Is there a bug in calculateSNPeff.pl?
    • What proportion of targeted SNPs are matches?
    • How many SNPs are present in target arms?
    • Are (heterozygous) mismatches more likely to occur in non-targeted regions?
    • What is the median coverage of targeted regions vs. non-targeted regions?

Comparing to 1KG[edit]

' All ' Targeted '
Match 16319 78.60% 15579 88.73%
Mismatch 596 2.87% 392 2.23%
Miss 3846 18.53% 1586 9.03%
  • More matches among targeted SNPs
  • Threshold coverage:
../Scripts/threshold.pl snp.targeted.1k 8 > cov8.target
../Scripts/threshold.pl snp.1k 8 > cov8.all  
' All ' Targeted '
Match 13548 92.11% 13274 92.78%
Mismatch 125 0.85% 111 0.78%
Miss 1035 7.04% 922 6.44%
Total 14708 14307
    • CONCLUSION: if SNPs are thresholded at 8x coverage, the additional exclusion of off-target SNPs does not significantly decrease the false positive rate.

SNPs in Arms[edit]

  • Worked on snpInArm.pl
../Scripts/snpInArm.pl orig.snp > snp.arms 
../Scripts/compare1K.pl snp.arms>snp.arms.1k
' Arms ' Arms >=8x '
Match 469 64.16% 409 82.96%
Mismatch 20 2.74% 12 2.43%
Miss 242 33.11% 72 14.60%
Total 731 493
  • Which of these are covered by other probes?
../Scripts/snpTargeted.pl snp.arms.1k >arms.target &
../Scripts/threshold.pl arms.target 8 >cov8.arms 
' Targeted ' Only arms ' Targeted (8x) ' Only arms (8x) '
Match 458 79.51% 11 7.10% 405 82.15% 4 20.00%
Mismatch 19 3.30% 1 0.65% 11 2.23% 1 5.00%
Miss 99 17.19% 143 92.26% 57 11.56% 15 75.00%
Total 576 155 493 20
  • But what about SNPs that are targeted but aren't in arms?
../Scripts/snpInArm.pl snp.targeted.1k > snp.targeted.arms &
' No Arms ' >=8x '
Match 15120 89.05% 12868 93.02%
Mismatch 373 2.20% 100 0.72%
Miss 1487 8.76% 865 6.25%
Total 16980 13833

ROC[edit]

  • Summary table:
Filter # of SNPs Miss Mismatches Total Errors
None 20761 18.53% 2.87% 21.40%
>=8x coverage 14708 7.04% 0.85% 7.89%
Targeted 17557 9.03% 2.23% 11.27%
Targeted (8x) 14307 6.44% 0.78% 7.22%
Targeted, no arms 16980 8.76% 2.20% 10.95%
Targeted, no arms (8x) 13833 6.25% 0.72% 6.98%