What Coverage allele-fraction threshold to use?
I UV mutated a haploid algae genome and want to view the variants in the bam file on IGV. What Coverage allele-fraction threshold should I use to look at variants?
• 2,518 views
•
link
updated
by
Ram
•
written
by
kristina.mahan
•
written
by
kristina.mahan
0 answers
No answers yet.
Log in to answer this question.
More posts like this
-
File format to view overlapping peaks on multiple separate tracks in IGV
written by Anisur Rahman •
-
How to control for UV mutational signature? (for recurrent somatic mutation)
written by ConvolutedGenome •
-
Variant Allele frequency
written by Maryam •
-
Hide small indels from PacBio long read sequencing on IGV
written by jan
-
Calculate the area covered more than a certain depth
written by Floydian_slip
-
Plotting Allele Frequencies
written by kristina.mahan
-
IGV: how is "Quality weight allele fraction" calculated?
written by omer.k •
-
Sub sampling bam files (dilution)
written by Chadi Saad
-
What is the coverage when allele frequency is < 1?
written by izaak.coleman1 •
-
Why a true variant is not getting called by Haplotypecaller.
written by AshishKS •
Depends on your sequencing coverage and desired False Discovery Rate.
The sequencing coverage is >200x
At each genomic position you have a specific coverage, say X. Y out of X reads may support an alternative variant, X - Y - reference. You need to perform a statistical test (there are tens of types of stat tests, more and less sophisticated) how unusual is to see Y reads given the error rate of your sequencing machine of Z. You can simply use a Binomial test. Then you get a bunch of p-values. You put these p-values into some FDR correction procedure and get your approximate threshold.
This procedure has its drawbacks and is not exactly correct, but may be useful.
Can you point me to some statistical tests?
For example, your probability of error as a substitution is 0.1% (in Illumina machines it is a very small number). You see 5 nucleotides A and 195 nucleotides B at some position. You apply Binomial test and find a p-value ( https://en.wikipedia.org/wiki/Binomial_test )
This is the simplest test, then people model errors with more efficient regression models with various link functions, but you may start with Binomial.