Differences

This shows you the differences between two versions of the page.

--- archive:bioinformatic_tools:bwa [2011/05/20 18:48]
svohr [After SeqPrep]
+++ archive:bioinformatic_tools:bwa [2015/09/04 09:06] (current)
68.180.228.52 ↷ Links adapted because of a move operation
@@ Line 29: / Line 29: @@
 </code>
+Note that BWA does seem to accept gzipped files, so there is no need to ungzip the read files, though the documentation doesn't mention this.
 ===== Quirks =====
 The SAM formatted alignments include a column labeled "inferred insert length" by the BWA manual, but in the SAM specification it is described as the "template length" or distance between the leftmost mapped base to the rightmost mapped base. The second description seems to
@@ Line 62: / Line 63: @@
 BWA was used to estimate the distribution of insert sizes in the Illumina runs for banana slug. The 454 reads were used as the reference and the Illumina reads were mapped onto them. The distribution of the insert lengths can be inferred from the pairs that map onto the same 454 read. This is possible because our insert sizes are smaller than the size of the 454 reads.
-Here is the frequencies of each inferred insert length from the SAM file from the paired end alignments for Illumina run 2. The mean inferred insert size for the barcode 7 reads is 258 bases and 138 bases for the barcode 8 reads.  Note that this differs considerably from the estimates of 411 bp for barcode 7 and 372bp for barcode 8 from the [[computer_resources:data|computer_resources:data]] page, which was based on bioanalyzer results for the DNA library.  What is the discrepancy? Is it different definitions of the length (including neither, one, or both reads in the length)? Why does the barcode 8 graph cut off so abruptly? (overlapping reads?) If the "inferred insert length" here is between the reads, then we need to add 200 for the read lengths to get the full DNA length, giving 458 and 338, which are fairly close to numbers reported by the bioanalyzer, but that would not explain the cutoff at 100.  If the inferred insert length here is the difference in the start positions in the same strand of the two reads, we would have to add 100 for the read length, getting 354 and 238, which seem a bit low.
+Here is the frequencies of each inferred insert length from the SAM file from the paired end alignments for Illumina run 2. The mean inferred insert size for the barcode 7 reads is 258 bases and 138 bases for the barcode 8 reads.  Note that this differs considerably from the estimates of 411 bp for barcode 7 and 372bp for barcode 8 from the [[archive:computer_resources:data|computer_resources:data]] page, which was based on bioanalyzer results for the DNA library.  What is the discrepancy? The difference is that the Bioanalyzer results include the adapters, not just the DNA that is sequenced, so the difference is actually fairly small.
+Why does the barcode 8 graph cut off so abruptly? (overlapping reads?)
 {{:bioinformatic_tools:run2_insert_size_histogram.png|}}
@@ Line 69: / Line 72: @@
 ===== After SeqPrep =====
-We ran [[bioinformatic_tools:seqprep|SeqPrep]] on run 2 to remove the Illumina adapter sequences and merge pairs that overlapped and mapped the remaining pairs to the 454 reference. SeqPrep removed most of the barcode 8 pairs that were mapped previously, but left most of the barcode 7 pairs that previously mapped unchanged.
+We ran [[archive:bioinformatic_tools:seqprep|SeqPrep]] on run 2 to remove the Illumina adapter sequences and merge pairs that overlapped and mapped the remaining pairs to the 454 reference. SeqPrep removed most of the barcode 8 pairs that were mapped previously, but left most of the barcode 7 pairs that previously mapped unchanged.
 {{:bioinformatic_tools:run2_seqprep_template_size_histogram.png|}}
-These histograms show the mapped lengths for the paired-end templates and the lengths of merged reads from SeqPrep along with the 454 read length distribution for comparison. In each of these, we can see the distinct range for the SeqPrep merged reads and the split between merged and unmerged pairs. Lengths less than 90 are probably incorrect. The higher frequency of these in run 1 can be explained its higher coverage.
-{{:bioinformatic_tools:run1_seqprep_histogram.png|}}
+The next histogram shows the 454 length distribution with the SeqPrep merged read lengths and the mapped lengths for the paired-end templates. From this we can see that the lengths of the 454 reads are greater than the illumina templates and the short templates lengths are not due to the lengths of the sequences in the 454 reference.
+{{:bioinformatic_tools:run_all_illumina_v_454_histogram.png|}}
+These histograms show the mapped lengths for the paired-end templates and the lengths of merged reads from SeqPrep. In each of these, we can see the distinct range for the SeqPrep merged reads and the split between merged and unmerged pairs. The counts for the paired-end reads are lower than the counts for merged reads because the paired-end reads had to map to one of the 454 reads. We can see that if the paired-end counts were increased by a factor of 10 (for the 0.1x coverage in the 454 reads) the merged and paired-end reads would form a continuous distribution. Lengths less than 90 may be incorrect. The higher frequency of these in run 1 can be explained its higher coverage.
+In the merged lengths for both run 1 and run 2 barcode 8 there is a gap of 10 lengths (66-75 for run 1, 105-114 for run 2 bc08) where no reads were observed. This is an artifact of SeqPrep's two methods for merging reads and the 10 base overlap requirement for merging.
+{{:bioinformatic_tools:run1_seqprep_histogram_r2.png|}}
+{{:bioinformatic_tools:run2_bc07_seqprep_histogram_r2.png|}}
+{{:bioinformatic_tools:run2_bc08_seqprep_histogram_r2.png|}}
+===== After Quake =====
+This process was repeated on the data after Quake correction. In addition to the paired reads and merge reads, Quake separates the reads whose pair could not be successfully corrected.
+{{:bioinformatic_tools:run1_quake_histogram.png|}}
+{{:bioinformatic_tools:run2_bc07_quake_histogram.png|}}
+{{:bioinformatic_tools:run2_bc08_quake_histogram.png|}}
+==== Mean Lengths ====
+Run 1
+| Template | 151.7 |
+| Merged   | 113.3 |
+| Single   | 65.7  |
+Run 2 bc07
+| Template | 250.3 |
+| Merged   | 142.5 |
+| Single   |  88.8 |
+Run 2 bc08
+| Template |  94.5 |
+| Merged   | 104.2 |
+| Single   |  61.3 |
+===== After Assembly =====
+This process was repeated using the [[archive:bioinformatic_tools:soapdenovo|SOAPdenovo]] contigs from ''assemblies/slug/SOAPdenovo-assembly2/k47_w_454_contigs'' instead of the 454 reads as the reference. The shapes of these distributions follow the same patterns as the ones found using the 454 reads, although there are more mapped pairs because of the higher coverage of the contigs and we see some longer templates sizes than before. For the run1 and run2 barcode 7 pairs, the results involve some self-reference, since the previous estimates were used to build the contigs to which we mapped the pairs. However, the pairs for barcode 8 were not used in the assembly because of their negative mean
+insert size (most pairs overlap but not in a way that SeqPrep could merge). When these were mapped to the contigs, we saw the same pattern as before but with more reads that mapped successfully.
+{{:bioinformatic_tools:run1_assembly_histogram.png|}}
+{{:bioinformatic_tools:run2_bc07_assembly_histogram.png|}}
+{{:bioinformatic_tools:run2_bc08_assembly_histogram.png|}}
-{{:bioinformatic_tools:run2_bc07_seqprep_histogram.png|}}
-{{:bioinformatic_tools:run2_bc08_seqprep_histogram.png|}}

Banana Slug Genomics

User Tools

Site Tools

Differences

Page Tools