I am working on paired-end Illumina NGS data, and the steps go like this: Assembly, demultiplexing, and the downstream data processing. Coverage (aka total read counts) is an important parameter to understand whether the generated sequence is a true sequence or due to a sequencing error. For DNA sequencing, what total read count threshold would you usually use (aka which reads would you discard?) when processing your NGS data? I am currently using 50x as my threshold. It may be a bit stringent, hence I am curious about alternative threshold values.
More on the terminology here: https://genohub.com/next-generation-sequencing-guide/
I asked the very same question to an Illumina specialist and he said 1x is enough to say a position is ON, although we used 10x for our RNA-seq analyses.
30x for germline mutations, 1000x for somatic mutations.
Mutation threshold for somatic mutations should be on 2%.
Next to 20.000 genes present in the human genome there are also 12.000 pseudogenes present from which about 50% (6000) pseudogenes have more than 90% sequence homology. You should be aware of this when looking at variants, due to misaligment.
I usually require 10 reads to support the alternative before I am investigate in any mutations. This way should be robust to seq error problems, since those should be more or less randomly distributed across the genome.
50x sounds for me super conservative, so you might loose a lot of heterozygous SNPs.
We evaluate the quality of data and usually a 20x is considered sufficient for detecting germline mutations except for genes that may be involved in mosaicism. In this case the value must be over 1000x
You already mentioned Genohub. There is also another interesting document which is probably answering your question and providing links to literature.
https://genohub.com/recommended-sequencing-coverage-by-application/
- Badges
- Science topic
- Similar topics
- Clinical Pharmacology
- Dosing