Matt:LabNotes/2013-7-26: Difference between revisions

From ZhangLabWiki
Jump to navigation Jump to search
>Mzcai
>Mzcai
Line 28: Line 28:


==Analysis of HL155: Representation Bias of Agi26k Oligos (Corrected)==
==Analysis of HL155: Representation Bias of Agi26k Oligos (Corrected)==
*WAS USING THE WRONG END OF PROBES AS REFERENCE WHICH WAS WHY ALIGNMENT RATE WAS < 3% (see hidden comments for that analysis)
**Create new fasta file with [[File:Probelist_to_fasta_revcomp.txt|Probelist_to_fasta_revcomp.pl]]
*Try both full reference (Agi26k_corrected.fa) and split between 0gap and 20gap (Agi26k0gap_corrected.fa & Agi26k20gap_corrected.fa)
===Alignment of Agi26k oligo reads to full reference===
*~33% alignment rate for 0gap expected because ~2/3 Indx10 of the reads are Dan's probes
*~33% alignment rate for 20gap expected because ~2/3 Indx12 of the reads are Dan's probes
*Using the same reasoning as for the CA12k Oligo alignment, redid alignment without "-k 1"
<!--*Using the same reasoning as for the CA12k Oligo alignment, redid alignment without "-k 1"
<!--*Using the same reasoning as for the CA12k Oligo alignment, redid alignment without "-k 1"
*Also separated 0gap and 20gap probes into two reference fasta files to prevent incorrect alignments
*Also separated 0gap and 20gap probes into two reference fasta files to prevent incorrect alignments

Revision as of 00:20, 27 July 2013

Analysis of HL152: Representation Bias of CA12k Oligos (Corrected)

  • I realized one of the parameters I have been using for bowtie2 ("-k 1") causes it to report the first good alignment but doesn't search for the best alignment
  • The other parameters that I changed from default (mismatch penalty, gap penalties) decreased the penalties from default allowing more alignments to be considered
    • The penalty for a single base gap is now equal to the penalty for single base mismatch while default a gap is about twice as high
    • This is okay for my purposes here because oligo synthesis introduces high rate of gaps but I still want to count those as aligned reads
    • I also want to consider as many reads as possible because I've shortened the reference to be very "specific"/defined (Not sure if this makes sense, I'll try to reword/explain better)

Using the 50bp segments of the probe reference because it will create the fewest false alignments (aligning to full probe sequence resulted in ~10% more alignments):

  • Low alignment rate expected because did not use Illumina index so the reads included any junk reads without an index
 /home/kunzhang/softwares/bowtie2-latest/bowtie2 --phred64 --mp 1,0 --rdg 0,1 --rfg 0,1 -x /home/mzcai/InSitu_HL152_130524_Analysis/CAprobes_50bp -q /home/mzcai/InSitu_HL152_130524_Analysis/s_2_1_unassigned.txt > /home/mzcai/InSitu_HL152_130524_Analysis/Readsalign2probes_CA.txt &
 mzcai@genome-miner:~/InSitu_HL152_130524_Analysis$ 29716630 reads; of these:
 29716630 (100.00%) were unpaired; of these:
   22676072 (76.31%) aligned 0 times
   1719956 (5.79%) aligned exactly 1 time
   5320602 (17.90%) aligned >1 times
 23.69% overall alignment rate

Counted up the reads for each probe using Count_reads_from_bowtie2_SAM_quantifyOligo.pl: CountofCAOligosv4.txt

  • Some variation is to be expected even for a perfectly uniform starting oligo pool because PCR is a stochastic process and each cycle will double some oligos and not others
  • Majority of probes have between 100-1000 reads
    • 747 probes with >1000 reads
    • 8 probes with <100 reads

File:RankOrderofCA12kOligoCounts.JPG

File:BinsofCA12kOligoCounts.JPG

Analysis of HL155: Representation Bias of Agi26k Oligos (Corrected)

  • WAS USING THE WRONG END OF PROBES AS REFERENCE WHICH WAS WHY ALIGNMENT RATE WAS < 3% (see hidden comments for that analysis)
  • Try both full reference (Agi26k_corrected.fa) and split between 0gap and 20gap (Agi26k0gap_corrected.fa & Agi26k20gap_corrected.fa)

Alignment of Agi26k oligo reads to full reference

  • ~33% alignment rate for 0gap expected because ~2/3 Indx10 of the reads are Dan's probes
  • ~33% alignment rate for 20gap expected because ~2/3 Indx12 of the reads are Dan's probes
  • Using the same reasoning as for the CA12k Oligo alignment, redid alignment without "-k 1"