Athurva Gore/LabNotes/2009-8-25
Jump to navigation
Jump to search
[edit]
ProbeSet Generation[edit]
- Currently creating TDMR set and CpGSNP set with new no-library-construction protocol
- This requires a constraint of arms+target = 105
- This will allow reads of 60-80 bp (after digestion) to cover the entire fragment, greatly improving the capturing protocol (as it seems that library construction is a major source of headaches)
NA12878 False Positives[edit]
- Noticed that many misses have high number of start/end sites for reads
- Checked actual read mapping; it seemed that the end of reads was causing the most trouble
- Many reads had several SNPs in the last few bases
- After discussing with Dr. Zhang, decided to test mapping and SNP calling using the following protocol.
- Reads were inspected after mapping to see if any SNPs were present in bases 36-40
- If any SNPs were present, the entire read was thrown out
- Hypothesis: Junk is getting ligated onto the ends of reads during library construction
- If we remove the junk reads, we should be able to greatly improve the false positive rate
- Remapped and SNP-called the HL006 s_4 (size-selected) and s_5 (not size-selected) sets
- Filtered out clonal reads
- Removed reads that had SNPs in locations >= the 36th base
- Wanted to see if false positive rate went down (though since entire reads are being rejected, coverage will decrease and therefore false negative rate will go up)
- Results:
- Stored in /media/TmpStore1/AG_scratch/Exome/NA12878/HL006/noclonal/newtest
- POST SPECIFIC RESULTS HERE
- Works extremely well!
- Number of non-dbSNP calls is reduced to ~100
- Lose ~1000 hits, but
- We only lose around 200,000 reads!
- Works extremely well.
- Will now apply it to IPS data.
IPS and Cancer[edit]
- Tested DF-6-9-9 and foreskin fibroblast non-sizeselected reads against each other (HL006, s1 and s3)
- However, removed clonal reads and reads with SNPs in locations >= 36th base as above
- Location: /media/TmpStore1/AG_scratch/Exome/DF-6-9-9/HL006/removeBadLig
- Running mapping/SNP calling scripts now
- Should reduce number of false positives
- False positive rate was found to be greatly decreased
- POST SPECIFIC RESULTS HERE
- Only a few SNPs not at dbSNP locations compared to before
- Need to investigate thoroughly, filter out bad calls
- Now running script to generate variant pairs between S3 and S1; looking for a gain of heterozygosity with particular interest