News and Announcements » |
Description:
Contains code for assigning taxonomy, using several techniques.
Given a set of sequences, assign_taxonomy attempts to assign the taxonomy of each sequence. Currently there are two methods implemented: assignment with BLAST and assignment with the RDP classifier. The output of this step is a mapping of input sequence identifiers (1st column of output file) to taxonomy (2nd column) and quality score (3rd column). The sequence identifier of the best BLAST hit is also included if the blast method is used (4th column).
Usage: assign_taxonomy.py [options]
Input Arguments:
Note
[REQUIRED]
[OPTIONAL]
Output:
The consensus taxonomy assignment implemented here is the most detailed lineage description shared by 90% or more of the sequences within the OTU (this level of agreement can be adjusted by the user). The full lineage information for each sequence is one of the output files of the analysis. In addition, a conflict file records cases in which a phylum-level taxonomy assignment disagreement exists within an OTU (such instances are rare and can reflect sequence misclassification within the greengenes database).
Example reference data sets and id_to_taxonomy maps can be found in the Greengenes OTUs. To get the latest build of those click the “Most recent Greengenes OTUs” link on the top right of http://blog.qiime.org. After downloading and unzipping you can use the following following files as -r and -t. As of this writing the latest build was gg_otus_4feb2011, but that portion of path to these files will change with future builds. Modify these paths accordining when calling %prog.
-r gg_otus_4feb2011/rep_set/gg_97_otus_4feb2011.fasta -t gg_otus_4feb2011/taxonomies/greengenes_tax_rdp_train.txt (best for retraining the RDP classifier) -t gg_otus_4feb2011/taxonomies/greengenes_tax.txt (best for BLAST taxonomy assignment)
Sample Assignment with BLAST:
Taxonomy assignments are made by searching input sequences against a blast database of pre-assigned reference sequences. If a satisfactory match is found, the reference assignment is given to the input sequence. This method does not take the hierarchical structure of the taxonomy into account, but it is very fast and flexible. If a file of reference sequences is provided, a temporary blast database is built on-the-fly. The quality scores assigned by the BLAST taxonomy assigner are e-values.
To assign the sequences to the representative sequence set, using a reference set of sequences and a taxonomy to id assignment text file, where the results are output to default directory “blast_assigned_taxonomy”, you can run the following command:
assign_taxonomy.py -i repr_set_seqs.fasta -r ref_seq_set.fna -t id_to_taxonomy.txt
Optionally, the user could changed the E-value (“-e”), using the following command:
assign_taxonomy.py -i repr_set_seqs.fasta -r ref_seq_set.fna -t id_to_taxonomy.txt -e 0.01
Assignment with the RDP Classifier:
The RDP Classifier program (Wang, Garrity, Tiedje, & Cole, 2007) assigns taxonomies by matching sequence segments of length 8 to a database of previously assigned sequences. It uses a naive bayesian algorithm, which means that for each potential assignment, it attempts to calculate the probability of the observed matches, assuming that the assignment is correct and that the sequence segments are completely independent. The RDP Classifier is distributed with a pre-built database of assigned sequence, which is used by default. The quality scores provided by the RDP classifier are confidence values.
To assign the representative sequence set, where the output directory is “rdp_assigned_taxonomy”, the you can run the following command:
assign_taxonomy.py -i repr_set_seqs.fasta -m rdp
Alternatively, the user could change the minimum confidence score (“-c”), using the following command:
assign_taxonomy.py -i repr_set_seqs.fasta -m rdp -c 0.85
Note: If a reference set of sequences and taxonomy to id assignment file are provided, the script will use them to generate a new training dataset for the RDP Classifier on-the-fly. Due to limitations in the generation of a training set, each provided assignment must contain exactly 6 taxa in the following order: domain (level=2), phylum (level=3), class (level=4), order (5), family (level=6), and genus (level=7). Additionally, each genus name must be unique, due to the internal algorithm used by the RDP Classifier.