As already answered, amplicon sequencing (metagenetic) and metagenomic present several bias ranging from the DNA extraction (which is shared by any molecular methods), genes targeted (for amplicons seq), sequencing depth, sequencing accuracy that will affect the results. Because we don't sequence everything in environmental samples (especially in soil), by definition sequencing is not quantitative by any mean. In other word, the number of sequences you get for a specific taxa is unlikely to be close to the real number of that taxa in your sample. If you want to quantify specific genes, the way to do it is to use quantitative PCR (which also have some bias).

However, some methods were developed trying to use amplicon sequencing as quantitative by adding a standard to your samples before DNA extraction: See the article from Smets et al 2016: Article A method for simultaneous measurement of soil bacterial abun...

So the question is what is the meaning of the number you get for each taxa? Regarding rarefaction, this is a slightly different issue that I will no answer here, there is quite few articles on it now; see McMurdie and Holmes: Article Waste Not, Want Not: Why Rarefying Microbiome Data Is Inadmissible

So is next gen seq semi-quantitative? Yes it is! Because you can compare within and across samples the difference in sequence numbers, the relative abundance, and how it differ. What matter is not the actual number but the difference between taxa within and between samples. The actual number is meaning less in the sense that it does not reflect the actual number in your samples. So with a deeper sequencing or different extraction method you might end up with the same number or 10 times more sequences for the same taxa. Where I would differ in my reply with Igor, is that the effect of the bias are more likely to be present between samples than within a sample because for a specific sample you reduce the bias of DNA extraction, as you work on the same DNA, and work on the same seq depth. Comparing between samples will be more affected by the effect of DNA extraction, PCR inhibitors, sequencing depth etc. So if in sample A you have 2000 seq of a taxa and 3000 in sample B, with a new extraction and sequencing you might have a different result...

Overall, in microbial ecology, sequencing data are always considered as semi-quantitative as soon as you do any analysis that include the number of seq, such as with PCoA, PERMANOVA, network, or a simple barplot to compare taxa and samples. The dominant taxa should remain dominant even after a different extraction and sequencing, however, difference between samples will differ to some extant, but you should find similar trend at a general level such as difference between treatment for example.

To conclude, we can also consider to some extant metagenetic/omic as semi-qualitative as we only sequence part of the diversity, we don't know the actually number of different bacteria, especially in highly complex samples...

Semi conservative replication means that each DNA strand will duplicate separately after they been away from each other during replication process and a new strand will be built on the separated strands as a template depending on base complementation

the attached images may explain the idea

GOOD LUCK

Dear Zahra,

Semi-quantitative and semi-conservative refer to two completely different things.

Hi Dietrich,

I think it has been quite well established that biases associated with DNA extraction, PCR and sequencing shift the abundance of taxa in metabarcoding analyses. There has been many benchmarking studies on that. When people say that sequencing is semi-quantitative, it mostly refers to the idea that these biases are likely consistent across samples. So it should be OK to use quantitative data, for example, to compare the abundance of a given taxa across samples, but not to compare different taxa within the same sample.

As already answered, amplicon sequencing (metagenetic) and metagenomic present several bias ranging from the DNA extraction (which is shared by any molecular methods), genes targeted (for amplicons seq), sequencing depth, sequencing accuracy that will affect the results. Because we don't sequence everything in environmental samples (especially in soil), by definition sequencing is not quantitative by any mean. In other word, the number of sequences you get for a specific taxa is unlikely to be close to the real number of that taxa in your sample. If you want to quantify specific genes, the way to do it is to use quantitative PCR (which also have some bias).

However, some methods were developed trying to use amplicon sequencing as quantitative by adding a standard to your samples before DNA extraction: See the article from Smets et al 2016: Article A method for simultaneous measurement of soil bacterial abun...

So the question is what is the meaning of the number you get for each taxa? Regarding rarefaction, this is a slightly different issue that I will no answer here, there is quite few articles on it now; see McMurdie and Holmes: Article Waste Not, Want Not: Why Rarefying Microbiome Data Is Inadmissible

So is next gen seq semi-quantitative? Yes it is! Because you can compare within and across samples the difference in sequence numbers, the relative abundance, and how it differ. What matter is not the actual number but the difference between taxa within and between samples. The actual number is meaning less in the sense that it does not reflect the actual number in your samples. So with a deeper sequencing or different extraction method you might end up with the same number or 10 times more sequences for the same taxa. Where I would differ in my reply with Igor, is that the effect of the bias are more likely to be present between samples than within a sample because for a specific sample you reduce the bias of DNA extraction, as you work on the same DNA, and work on the same seq depth. Comparing between samples will be more affected by the effect of DNA extraction, PCR inhibitors, sequencing depth etc. So if in sample A you have 2000 seq of a taxa and 3000 in sample B, with a new extraction and sequencing you might have a different result...

Overall, in microbial ecology, sequencing data are always considered as semi-quantitative as soon as you do any analysis that include the number of seq, such as with PCoA, PERMANOVA, network, or a simple barplot to compare taxa and samples. The dominant taxa should remain dominant even after a different extraction and sequencing, however, difference between samples will differ to some extant, but you should find similar trend at a general level such as difference between treatment for example.

To conclude, we can also consider to some extant metagenetic/omic as semi-qualitative as we only sequence part of the diversity, we don't know the actually number of different bacteria, especially in highly complex samples...

Aimeric,

Great summary, thanks for that! Just one comment: I'm still not convinced that comparing different taxa within a sample is benefited from reduced bias, as any DNA extraction protocol favours to some extent some taxa over others. The same is true for primer efficiencies. And on top of that you have other factors that greatly affect the observed abundances, such as genome size, number of copies of the target gene and so on. As a result, taxa with cell walls which are easier to lyse, with annealing regions matching perfectly the primer sequences and with more copies of the target gene will appear more abundant than they are and vice-versa. But the idea is that these biases are likely consistent across samples (providing that the same DNA and PCR experimental protocols have been applied). While biases will cause a certain taxa to be over-represented within a sample (thus making quantitative interpretations meaningless), this over-represented taxa will likely be over-represented in all samples in a similar fashion (thus allowing to some extent between-sample comparisons).

Hi everyone! such an interesting topic to think about! These "new" technologies are extremely useful but based on assumptions that arise in growing uncertainties. Form environmental samples, I also have my concerns about multi nuclear spores of fungi which can impact on the number of sequence reads. Moreover, we assume that each gene of different taxa produce a single copy during amplification but it might not be always true…. I think that abundance patterns from sequencing data involve many decisions that we should not only make but also faith!

Hi Igor,

I think it really depend of the environmental samples you are working on. Working on soil, I think that there is more variability between samples than within a sample. Because of the diversity of soil samples (clay, silt, sand, OM, density, pH, humic acid...) that strongly affect DNA extraction efficiency, with yield ranging from 5->100 ng/µl, this have huge effect on all the subsequent analysis (from PCR to sequencing depth) and is stronger than the variability within a single sample. So, the bias are not consistent across soil samples.

I acknowledge the fact that some species will be more difficult to be extracted than other. However, we can expect this to be more variable between samples, as the community composition vary between samples, and we don't know the ratio of microorganisms easy/difficult to extract. The number of gene copies per genome highly depend on the gene of interest, varying from 1-3 copies for some functional genes, to 5~15 copies for 16S. However, this is true within and between samples. The copies number and genome size have more effects on metagenomic than metagenetic. The question here is are these bias stronger than the biological variability, which is directly related to the nature of your samples

I think the dominant and rare taxa found in a soil sample are likely to be true even with a different extraction method or sequencing. However, taxa with similar seq number are more difficult to compare within and between samples.

At the end of the day, as soon as we do any analysis (PCoA, PERMANOVA, network...) you consider within and between samples analysis and we can't really avoid it. That's why biological replicates are important to compare taxa and effects of factors, and I still see too often non replicated studies using omics techniques...

Environmental samples, such as marine ecosystems are likely to have different constrain due to difference abundance and diversity, with less issue due to DNA extraction between samples than for soil.

Sorry for the long reply. Just a really interesting discussion/debate to have