After high-throughput sequenceing of 16S rDNA, the sequencing depths of different samples usually vary a lot. The sequencing depth can affect alpha and beta diversity analysis, therefore, we usually used the strategy of rarefaction (randomly sub-sampling of sequences from each sample) to equalize the number of sequences per sample. But when we performed functional genes' diversity (e.g. amoA gene of ammonia-oxidizing microorganisms), we often used a clone library method due to the limitation of read length of NGS. As a result, we only obtained very limited numbers of sequences (e.g. 50 to 100 sequences varied among samples) in each sample. If we randomly sub-sample like the 16S rDNA data, we may lost nearly half of the sequence number in some samples and this should have great influence on the alpha or beta diversity. So, in this case, if we can calculate the alpha and beta diversity based on the relative abundance data of OTU? i.e. before calculating the diversity, the data in each sample were firstly unified through divided the total sequence number in each sample. Is this transformation is reliable and scientific? Is there anyone using this method to calculate alpha or beta diversity? If you have related references, I will very appreciate that.
Thank you!
I don't think it will be a very good idea. Firstly you have very limited number of sequences 15-100 is nothing in metagenomics (where now a days genomics community works on millions of reads) and secondly 16S data is not very good way to predict functions. metagenomics (shotgun) or metatranscriptomics can provide you with the details of functional diversity. Hope this helps.
Dear Mitra,
Thank you for your answer. It's correct that we have limited number of suquences, and this may be a disadvantage of clone library method, but we can obtain longer sequence length using clone library. Furthermore, I think if we want to study functional gene diversity, it's better to unified the sequences' number or coverage. Of course, if we have enough money, metagenomics would be a very good choice. Thanks again.
regards,
Guangshan
Hi Guangshan,
I agree with Mitra, metagenomics/metatranscriptomics would provide would be better for functional comparison. However, my point is to warn you against the use of relative abundance for alpha/beta diversity evaluation and comparisons. It would be misleading if not wrong. The reason is that it assumes that all the components (bacteria) present in the samples are known which it is never the case (even at vey high sequencing depth). In other words, it will reduce the composition of each sample to what it is known to be in each of them, excluding the possibility that more species could be (are given the low coverage used!!!!!) in the sample. This is the field of compositional analysis, for more info you can look at:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4030730/
Hope it helps
Luca
Hi Luca,
Thank you very much for your answer. I got it, I think you may mean that the relative abundance based diversity also could not solve the low coverage problem in some samples. In this case, it would be misleading if we compared the diversity indexes among different samples.
Regards,
Guangshan
- Badges
- Science topic
- Similar topics
- Biological Science
- Immunology
- Antibodies