Daniel:Notebook/GenomeMiner/2013-9-25

From ZhangLabWiki
Jump to navigation Jump to search

Mock HL155 (Started 9/9/2013)[edit]

Back to Calendar

Mimic Data Using Short Reference (BWA)[edit]

Error Counting Results[edit]

Error Rate: 1.177%
Error Rate of Insertions: 0.195%
Error Rate of Deletions: 1.42e-06%
Error Rate of Substitutions: 0.982%

Overview of Recent Results[edit]

Comparing Real Data[edit]

File:ProbeComparison Improved.png|File:AgilentMYCA ErrorComp090313.png

Left: Uses full length reference file; Right: uses shortened reference files

Comparing Bowtie and BWA Reference File Length Data[edit]

Name Pct Aligned Subtitution Rate Insertion Rate Deletion Rate Substitution Accuracy Insertion Accuracy Deletion Accuracy
Mimic Data (Full Ref) BWA 97.84 1.017 0.192 0.034 99.3 73.8 68.0
Mimic Data (Short Ref) BWA 95.44 0.982 0.195 1.43E-06 97.2 75.0 0.0
Mimic Data (Full Ref) Bowtie 95.47 0.887 0.195 5.52E-04 87.8 75.0 1.1
Mimic Data (Short Ref) Bowtie 96.4 0.91 0.188 0.028 90.1 72.3 56.0
Actual NA 1.01 0.26 0.05 NA NA NA

File:Bowtie-BWA-Compare-092513.png

Discussion[edit]

From above, we can see that both Bowtie and BWA do not count deletions accurately without the given reference sequence. So, looking back at the original data, it now makes sense that using the shortened reference sequence had fewer errors. At the time, we thought it was good that the error decreased. However, the majority of the errors lost were deletions, and the mock sequence data has shown that deletions are seriously undercounted using the shortened reference sequences. It's likely, therefore, that the original error counting was more accurate.