Hello! I have bacterial (16s V3-V4) and fungal (ITS2) miseq data from soils collected on different farms where I'm hoping to compare microbial networks in different farming types (conventional and organic). While OTUs have traditionally been used for network analyses, I'm wondering if, with the push for ASVs (https://www.nature.com/articles/ismej2017119), ASVs should be used instead? From the cautionary blog posts from Noah Fierer's lab (http://fiererlab.org/2017/10/09/intragenomic-heterogeneity-and-its-implications-for-esvs/), I'm a bit hesitant to use ASVs in network analyses due to potential, strong positive correlations being the result of simply having multiple sequence variants from the same organism. The only published example I've found with networks of ASVs (https://www.nature.com/articles/ismej201729) evaluates ASVs within OTUs, not just ASVs. Should only OTUs be used in the analysis of microbial networks? Or can ASVs be used instead, and the lack of published networks with ASVs simply due to the recent development of programs that produce ASVs? Many thanks for thoughts and ideas on this.
Not direct answer, but still might be helpful
* We we want to answer about "What is an OTU?", I guess we dont have any clear answer to it. Or they are just a sequence originated by clustering of various sequences at different clustering threshold. Thus, they are random and not a factual unit of taxonomy.
* If they are not a factual unit, how can they give factually correct answer to any associated question?
* OTUs have traditionally (as you describe) been used, because there were no other alternatives and all algorithm resulted in OTUs. Due to the development of improved algorithm like DADA2, which can resolve variation at even single nucleotide levels, OTU definition is fading out now. Thus, use of ASV is promoted.
Hello, Identifying Operational taxonomic units from raw sequences used clustering approach. Using 97% identity cut-off was a standard approach and often closed reference OTU picking was accepted in the sicentific community. During the time of the development of tools and bioinformatics approaches this was possibly the best available method. However, as with many other fields of science, the knowledge has been updated. Evolution of bioinformatics approaches is a constant process. Based on current knowledge, the cons of 97% OTU picking strategy.
For more info see link below.
Good luck !
https://mibwurrepo.github.io/Microbial-bioinformatics-introductory-course-Material-2018/set-up-and-pre-processing.html
I agree with Dr. Darbinyan, I think, due to lack of published networks with ASVs, OTU picking strategy is the best option to study microbial network.
There are already articles about ASVs and fungi (see below). When you search with google type also exact sequence variants (ESVs) instead of ASVs and you find more results.
Here some suggestion and interesting articles about ASVs:
Article in Fungal Ecology comparing ASVs and OTUs - ITS1F/ITS2 primers
Bioinformatics matters: The accuracy of plant and soil fungal community data is highly dependent on the metabarcoding pipeline
https://www.sciencedirect.com/science/article/pii/S1754504818302800
Article about ASVs in SCIENCE – Fungi ITS2
Fungal diversity regulates plant-soil feedbacks in temperate grassland
http://advances.sciencemag.org/content/4/11/eaau4578/tab-pdf
Article ASVs (ESVs) and OTUs - ASM Journal– ITS2 -
Broadscale Ecological Patterns Are Robust to Use of Exact Sequence Variants versus Operational Taxonomic Units
https://msphere.asm.org/content/msph/3/4/e00148-18.full.pdf
Article about ASVs in Frontiers in microbiology–Fungi ITS2
Fungal Biodiversity of the Most Common Types of Polish Soil in a Long-Term Microplot Experiment
https://www.frontiersin.org/articles/10.3389/fmicb.2019.00006/full
Super interesting discussion and comments about ASVs and OTUs-
http://fiererlab.org/2017/05/02/lumping-versus-splitting-is-it-time-for-microbial-ecologists-to-abandon-otus/
I think to some extent this is a semantic debate in that we could operationally define the ASV or ESV as our taxonomic unit (making it an OTU). My key point though would be that clustering at a 97% similarity doesn't deal with the problem of multiple counting that you are concerned about. You would still be counting each instance of a gene, even if a cell has multiple copies. But the downside would be that you would be bundling several genes from potentially unrelated organisms into a single OTU. As it is ASVs of the 16S v3-v4 are still sometimes not specific enough to resolve strain-level differences. So I think maintaining information is usually better, and so even if ASVs aren't perfect, they're definitely better to us than arbitrarily clustered OTUs.
I tend to spam ASVs whenever i can - their underlying theory looks more accurate than OTU clustering.
The reason is OTU clustering is arbitrary. You're assuming that just because sequences are somewhat similar or dissimilar to one another (like the 0.97 criteria commonly used for OTU picking) that they're potentially different organisms, with a technique whose resolution is often depending on one nucleotide. With ASVs, only sequences that are 100% identical to each other are considered the same species, so the confounding factor of nearby aligned sequences is essentially removed.
Choosing between ASVs and OTUs shouldn't be your only concern. Metabarcoding/amplicon analysis is filled with pitfalls. Robert Edgar (USEARCH author) has recently posted a series of short, yet extremely informative videos on 16S analysis that should give you a ton of support in making your own choices.
Links to the videos:
On 16S diversity: https://youtu.be/EwvMh4JcCWI
On 16S taxonomy: https://youtu.be/8vATfij_dWc
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics