Ultra deep sequencing does not make that much sense when we talk about finding genetic variation in humans. If you have a sequencing depth of about 20 (given the quality is fine) you are very certain that you call the correct genotype. The only problem is that if you have a mean depth of 20 you will have very many sites with lower depths and then you can not call genotype and you can not find rare variation. If only the sequencing-manufacturer could promise a coverage instead of mean depth. Example: It would be nice if the manufacturer could promise a min. coverage of 20X for 90% of the target region. Then I would be very certain that for 90 % of my target I would find all mutations. A mean depth, per see, does not tell you much. The only reason for increasing depth is to increase the overall coverage.

(This answer is only viewed from the perspectives of finding genetic human variation).

Hi Navonil. Apart from somatic mutations, I guess that ultra deep sequencing might be useful for the detection of mosaicisms. Besides, unsufficient sequencing depth leads to a bunch of false postives or even negatives. I already observed this for exomes with "basic" depth of 50-100X. Now, I don't know where to place the threshold. Which depth would be sufficient not to get too many artefacts?

I imagine the question applies to sequencing of the human genome. The value of ultra-deep sequencing is probably lost there, beyound the needed coverage to ensure reliability. It becomes really important, however, if you need to look for the source of pathogenicity beyond that, e.g. in metagenomics. There is a lot of progress being made in looking at the human microbiome, and considering its richness and much higher genetic variability, I very much believe that deep sequencing is essential to capture the complexity of the picture. If one wants to understand and attempt to profile a particular condition/disease in that context then greater depth and accuracy are quite important.

"Ultra Deep Sequencing" is required when you are looking for a very small change (of perhaps one gene) in a group of heterogeneous cells (cancer from a mixed population). I'm interpreting your question in relation to RNA-Seq. With the greater coverage (genome depth of >100x) it is possible to discover changes that may be diluted out or overshadowed by the more abundant changes; such as mutations on cell-cycle factors. If the "cancer" cells are derived from a homogeneous population, going deeper (more coverage) may uncover less obvious factors that may be novel. Hope this helps.

I totally agree with David. Actually, ultra deep sequencing is relevant when you are looking for a rare event. "Bypassing" this dilution issue thanks to a high coverage should enable for example to search variant alleles from tumour cells lost among overrepresented wild alleles from normal cells in circulating DNA. I didn't think of metagenomics, like Lila, because I assume that we rather see things through the prism of our own activity, but it's quite comparable. In the same way, ultra deep sequencing may be useful for the detection of viral DNA within a blood sample. In that, UDS is inot a mere technical performance, it has a biological relevance.

Ultra deep sequencing does not make that much sense when we talk about finding genetic variation in humans. If you have a sequencing depth of about 20 (given the quality is fine) you are very certain that you call the correct genotype. The only problem is that if you have a mean depth of 20 you will have very many sites with lower depths and then you can not call genotype and you can not find rare variation. If only the sequencing-manufacturer could promise a coverage instead of mean depth. Example: It would be nice if the manufacturer could promise a min. coverage of 20X for 90% of the target region. Then I would be very certain that for 90 % of my target I would find all mutations. A mean depth, per see, does not tell you much. The only reason for increasing depth is to increase the overall coverage.

(This answer is only viewed from the perspectives of finding genetic human variation).

Thomas, I understand your problematic and I agree with you that a mean depth doesn't mean much. Yet, as far as I am concerned, I would rather see deep sequencing as a sequencing targetting a single gene or a region for example, with a high coverage and a mean depth of 100-1000 and above. You don't need to reach such a depth for the identification of a novel variant which would be represented in 20-50% of your reads, because it would be heterozygous. However, in case of a rare event, i.e., a variant present in 1/100 or less of the total alleles of a given sample, or in case of a mosaicism, then it may be helpful to rely on ultra-deep sequencing. That's the way i see the things.