johnne / clean_asv_data Goto Github PK

Create script to generate

resolved taxonomic annotation for all clusters in the corresponding counts file.

I am suggesting a simple approach where we sort the cluster annotations according to the number of reads. By definition, our clusters will be uniformly annotated down to the family level. From the genus level, I suggest we accumulate annotations by read number until we reach some critical threshold, like 50%. If we have a single annotation then, we use that annotation for the cluster. Otherwise we use an annotation like “unresolved” to indicate that there is ambiguity in the annotation of the cluster at that level. I like to use a distinct term, because the majority-rule annotation may well be something with “_X” or “unclassified”.

For Arthropoda:

At 80%, there are 724, 1258 and 2269 unresolved clusters

Script to generate a file with:

all ASVs with the total number of reads and the total number of samples in which they occur (and possibly other stats).

Script to generate a file with:

all clusters and their read numbers in all samples

Script to generate a file with:

a subset file of ASVs after removing clusters/ASVs unclassified at the family level (“_X” or “unclassified”), clusters with < 3 reads, and ASVs occurring in more than 20% of blanks.