Currently, my analysis of the fungal data related to Vanilla species generated from Illumnia sequencing.
I used QIIME and UCLUST in the bioinformatics analysis. After completing the bioinformatics analysis, there generated a lot of sequences and OTUs in the data.
My colleague suggested that I use an R package called phyloseq to create a phyloseq object to begin my analysis. I used another R package as well, called decontam to detect contamination and supposedly remove them out of the data.
I potentially have identified 755 OTUs as fungal sequence signals. I was wondering that if you have through your works encountered or used a method to filter out the "noise" in the data. I.e. have you ever set up an abundance threshold for OUTs with fewer than 10 or 100 reads?
Have you ever encountered an instance were by looking at the OTUs that were identified as fungal sequences, how confident would you be to use this to data analysis to identify the orchid mycorrhizal fungi (OMF)?
I was thinking of drawing a comparison between populations using regression .
As well as using a PCoA and NMDS for the analysis. Any and all tips or methodological advice you could share on how to move the data to a statistical test format would be greatly appreciate.
Thank you for your attention, time, and guidance.
It is a very long question and somehow misses what you actually want to do.
I don't work with fungi however, the basic methods are same for the fungal or bacterial data analysis. Now, first question in your big question is the threshold. I don't understand what you exactly mean by noise. If you mean the sequence with bad quality, they can be removed by quality filtering and if you mean non target sequences, that must be removed by some other methods before clustering. Clustering must be done with the target sequences only. Therefore, removal of singletons and chimera sequences are perforemed prior to clustering.
Now it comes to clustering. Since, target sequences (denoised) are to be clustered, it is user specific or what is general clustering threshold (as in literature) for that target. Based on the threshold percentage, the minimum size of cluster can be defined. And this is user specific/variable.
I did not get the latter part of the question completely. However, I would not recommend to perform statistical analysis if you are not sure what do you have in your sequence data. Or do you want to perform some analysis as you said "regression" to find out Mycorrhizal fungi, you can try. I don't have idea what is the sequence variation in such fungal class.
Abhijeet Singh Thank you for the feedback, we are still going through a few protocols on this subject. More to follow.
- Badges
- Science topic
- Similar topics
- Mathematics