Hello everyone,
I have performed a 16S rRNA differential-abundance meta-analysis, and some taxa in my results are not at the genus level but only at the family or order level. Does anyone know of any tools or methods that can help me identify the corresponding genus level for these taxa?
Thank you!
You can try BLAST against a curated 16S database (e.g., SILVA) and assign genus using only high-confidence hits — e.g., ≥99% identity, ≥99% coverage, and E-value ≤ 1e-10, or simply keep the single best hit per sequence for a reliable genus-level match.
— Soyeon Kim
Dear @soyeon_kim
Thank you for your help.
Hi @iordanis , hi @soyeon_kim ,
I strongly discourage this, because very frequently there can be a tie for top hit, even from many different accessions (even from different genera). So only taking one can lead to misclassification.
The reason why some studies in your meta-analysis only classify some sequences at order or family level is probably because a classifier was used that attempts to find a consensus lineage of probable hits, e.g., using a last common ancestor approach. This indicates, e.g., that that sequence had hits to different species, genera, even families (in the case of order-level classifications), and so deeper classification was not possible. Of course, in a meta-analysis you should repeat this process to ensure that methodology is consistent between studies, but I would also recommend this same approach of using a consensus- or confidence-based classifier (such as those used in QIIME 2), do not just take the top blast hit.
Of course, I primarily recommend using tools within the QIIME environment. However, BLAST can serve as an optional approach for achieving higher resolution, though I recommend using it cautiously with carefully filtered results only. Thank you for sharing your insights