features bacterial retio is like in sample or not?

Hello!
I had a problem with memory while feature classifying, so I take an idea to blast feature by myself.
My question is could i interpreter my feature ratio as a ratio of components in my sample? For ex, could I say looking at the picture below that my sample consist of 76 procent lactobacillius?

Thanks a lot!

Hi!
Yes, you can. But actually the percentage of concrete taxonomy may increase since 76% is a ratio of one feature, but there may be more features assigned to the same taxa

1 Like

Thank you, Timur! Yes, I should blast all features and then make a taxonomic conclusion. Could I use results of such path in medical way, for ex, to analyse the existence of pathogens in patient samples for diagnosis?

I would try to find a stronger machine to assign the features to taxonomy with GG (greengenes) classifier to be more consistent. But I don’t know about how certain you may be about pathogens (I don’t have such experience). 16S amplicon data is good for exploring microbial communities up to Genus level. Is it good enough? Will wait here to hear from other participants.

1 Like

Agreed! However, depending on the size of the amplicon, it is sometimes possible to have species-level resolution with 16S sequencing. Not the case here, as per the figure provided by Irina, the amplicon is quite short (123nt), which is in general insufficient to detect specific pathogens. Even so, genus-level can still be useful to make conclusions like “we found a high relative abundance of genus A, which includes species X and Y, that are detrimental in this body site; to be confirmed if X and Y are really there”.

Anyway, also agree that it is essential to use an algorithm to assign features taxonomy, as manual classification is very time-consuming and prone to error. Besides, it would make all the downstream analyses much more direct and intuitive.

Hope the memory issue is solvable! Good luck

1 Like

Vitor, hello!
Thank you for recommends! Short aplicon length is a result of my second problem - small overlap (5-11 bp). Dada2 removed about 90 percent of reads because of it, so i gave a left read like single end to get at least something and dada2 removed only 20 percent, but I took short amplicons. I do not know how to solve this problem, whold be glad to hear any ideas

1 Like

I believe you already made the best choice when deciding to continue the analysis with single end reads. The best thing you can do now is to use one of the methods of the feature-classifier plugin to assign the taxonomy of your ASVs. After doing that, at least you can be confident that the taxa that you find are truly part of your samples. You will certainly not have high taxonomic resolution (like species-level), but even with full-length amplicons this is rarely achievable, so… :confused: Sorry I can help you any further

3 Likes