qunfengdong
commented
2 years ago Thanks for the report. Let us look into this and get back to you (perhaps in a few days). Re. "a minimum number of blast hits to use for a given OTU when aligning and estimating the taxonomy", it's a tricky thing that we had thought about it but did not figure out how to incorporate into our statistical framework. Imagine this case: your query only matches ONE database hit (just one, so it seems low), but it was a PERFECT match. I would still trust the result because it may simply reflect that there happens to be just one database entry from this taxon.
lplough
YJulyXing
Hi There, I am getting about 1/3 of my OTU assigned taxonomy (output) from BLCA (2/3 "unassigned"). In some cases, those OTU/sequences have poor or no match to the blast database provided and even doing a remote blast against the entire genbank nt database provides no match. However, there are a number of OTU/sequences that return many ( often up to requested # of hits, e.g. 50 by BLCA default) with relatively low e values (<1e-20) 70-80% alignment length, and 80-95% identity). These would seem like usable matches that BLCA could estimate LCA on, though of course I don't expect high confidence in species level assignments when %ID is lower than 95%.
Anyway, have tried to alter (relax substantially) some of the filter settings (arguments) in
2.blca_main.pyscript, specifically for what I think is the BLASTn part of the script, but still get many many 'unassigned' results for OTU that would appear to have decent blast matches. I am wondering if I am misunderstanding how to set these arguments or where in the pipeline the filters are being applied?For example, I am running this:
2.blca_main.py -i 97.centroids.fa -r Euk.Tax.BLCA.FINAL -q Euk.Co1.Tax -c .5 --iset 60Following the description of the options from BLCA (or
python 2.blca.main.py --help), I was trying to set the coverage filter pretty low (to 50%) and the minimum identity to 60% , also pretty low, to see if I could get more OTU to have some taxonomy information produced.Here are my results (top lines of the *.blca.out results file from that BLCA run:
This is fine, but when looking at the blastn output used in the BLCA pipeline, I am not understanding why e.g. OTU2 was not assigned ( 90% identity, coverage 71.3 %) but OTU 6, with just 3 hits, was?
I feel as though I am not understanidng some of the '2.main.blca.py' options or am not implementing them correctly? And, given that OTU6 has just 3 blast hits returned, Im not sure how much Id trust the LCA results here? Can (should we) we set a minimum number of blast hits to use for a given OTU when aligning and estimating the taxonomy?
Thanks for any input you can provide!
LP