Athurva Gore/LabNotes/2009-8-26: Difference between revisions
Jump to navigation
Jump to search
>Ajgore (New page: {{ AGLabEntry|2009-8-25|2009-8-27 }} =IPS and Cancer= * IPS data still had a large number of false positives, despite filtering * Upon looking at SNP distribution, found that bases after ...) |
>Ajgore |
||
(2 intermediate revisions by the same user not shown) | |||
Line 5: | Line 5: | ||
* Upon looking at SNP distribution, found that bases after #25 had very high false positive rate | * Upon looking at SNP distribution, found that bases after #25 had very high false positive rate | ||
* Reran filter, discarded reads with SNPs after base 25 | * Reran filter, discarded reads with SNPs after base 25 | ||
** IPS Set: 11314 in dbSNP, 2144 non-dbSNP --> 9370 in dbSNP, 1099 non-dbSNP | |||
** Fibroblast Set: 9254 in dbSNP, 2504 non-dbSNP --> 7355 in dbSNP, 1187 non-dbSNP | |||
* The number of non-dbSNP misses needs to be reduced further without removing more hits | |||
** How to accomplish this? | |||
*** Reduce quality of SNPs found towards the end of reads by a certain factor | |||
*** Trim reads when a SNP is found? | |||
*** Trim all reads and check data? | |||
*** Be skeptical of a train of SNPs? Perhaps throw out reads with more than two SNPs? | |||
=PGP High False Positive Data= | =PGP High False Positive Data= | ||
* | * Plotted SNP distributions for PGP1 and PGP2 data (06022009, Lanes 1 and 2) | ||
* Found a similar issue with SNP distributions as in the IPS sets | |||
* However, these used longer reads; errors seem to start around read 50 | |||
** '''INCLUDE GRAPHS OF SNPDIST HERE''' |
Latest revision as of 00:03, 27 August 2009
[edit]
IPS and Cancer[edit]
- IPS data still had a large number of false positives, despite filtering
- Upon looking at SNP distribution, found that bases after #25 had very high false positive rate
- Reran filter, discarded reads with SNPs after base 25
- IPS Set: 11314 in dbSNP, 2144 non-dbSNP --> 9370 in dbSNP, 1099 non-dbSNP
- Fibroblast Set: 9254 in dbSNP, 2504 non-dbSNP --> 7355 in dbSNP, 1187 non-dbSNP
- The number of non-dbSNP misses needs to be reduced further without removing more hits
- How to accomplish this?
- Reduce quality of SNPs found towards the end of reads by a certain factor
- Trim reads when a SNP is found?
- Trim all reads and check data?
- Be skeptical of a train of SNPs? Perhaps throw out reads with more than two SNPs?
- How to accomplish this?
PGP High False Positive Data[edit]
- Plotted SNP distributions for PGP1 and PGP2 data (06022009, Lanes 1 and 2)
- Found a similar issue with SNP distributions as in the IPS sets
- However, these used longer reads; errors seem to start around read 50
- INCLUDE GRAPHS OF SNPDIST HERE