| News and Announcements » |
This tutorial explains how to use the QIIME pipeline to analyze 18S, or mixed 18S/16S, data. While most of the steps are identical to the standard 16S pipeline described in the overview tutorial, reference data that include eukaryotic sequences may be required for OTU picking, taxonomic assignments, and template-based alignment building.
To this end, a QIIME-compatible version of the Silva 104 release, which contains data for all three domains of life, is available here: http://www.arb-silva.de/download/archive/qiime/. The Qiime_files_r104.zip archive contains several files that will be utilized in this tutorial, and can be used for analyzing 18S datasets. Notes about how these datasets and tree were constructed are listed at the bottom of this tutorial.
A sample fasta file containing sequences from archaea, bacteria, and eukaryotes can be found in the /qiime_tutorial/18S_tutorial_files/ which will be used along with the Silva reference set in this tutorial.
Each of the following steps are described in this tutorial (the remaining analyses are identical to those described in the general 16S tutorial):
Here we will cover picking OTUs. Using the default settings with pick_otus.py is no different than with 16S data. Sequences will be clustered at 97% identity and new clusters will be allowed. However, one can utilize a reference based approach with the Silva dataset listed above that discards sequences that do not cluster with the reference set. The advantage to using this approach is that one can use the reference taxonomic assignments and tree rather than generating new ones, however, a large portion of sequences may be discarded.
To pick OTUs against the Silva reference data set and discard any clusters that do not match, use the following command:
pick_otus.py -i 18S_tutorial_sample_seqs.fna -m uclust_ref -C -r QIIME_files/rep_set/silva_104_rep_set.fasta -o uclust_ref_picked_otus/
This will generate a 18S_tutorial_sample_seqs_otus.txt OTU mapping file in the uclust_ref_picked_otus folder. An OTU table with taxonomy strings from the reference dataset can be built directly. To do so now, use the following command:
make_otu_table.py -i uclust_ref_picked_otus/18S_tutorial_sample_seqs_otus.txt -t QIIME_files/taxonomy_mapping/Silva_taxa_mapping_104set_97_otus.txt -o otu_table_uclust_ref.txt
which will generate an OTU table, as shown below:
Note
To generate de novo OTUs, use the following command:
pick_otus.py -i 18S_tutorial_sample_seqs.fna -o uclust_picked_otus/
This will create a 18S_tutorial_sample_seqs_otus.txt file in the uclust_picked_otus directory, which can be used to make a representative sequence set for downstream analysis with the following command:
pick_rep_set.py -i uclust_picked_otus/18S_tutorial_sample_seqs_otus.txt -f 18S_tutorial_sample_seqs.fna -o rep_set.fna
To assign taxonomies to the de novo OTUs that were just generated, use the following command:
assign_taxonomy.py -i rep_set.fna -o rdp_assigned_taxonomy/ -t QIIME_files/taxonomy_mapping/Silva_RDP_taxa_mapping.txt -r QIIME_files/rep_set/silva_104_rep_set.fasta
If BLAST is the preferred method of assignment, the generic taxonomic mapping file can be used instead:
assign_taxonomy.py -i rep_set.fna -o blast_assigned_taxonomy/ -t QIIME_files/taxonomy_mapping/Silva_taxa_mapping_104set_97_otus.txt -r QIIME_files/rep_set/silva_104_rep_set.fasta -m blast
Finally, an OTU table can be built which includes the taxonomic assignments (in this case we will use the RDP assignments):
make_otu_table.py -i uclust_picked_otus/18S_tutorial_sample_seqs_otus.txt -t rdp_assigned_taxonomy/rep_set_tax_assignments.txt -o otu_table.txt
It may be desirable to split the OTU table according to domain for mixed 16S/18S datasets. To do this, we will use the split_otu_table_by_taxonomy.py module.
We will split the OTU table generated in the last step at the domain level, 2, by using the following command:
split_otu_table_by_taxonomy.py -i otu_table.txt -L 2 -o separated_otu_tables/
The output directory, separated_otu_tables, will contain an OTU table for archaea, bacteria, and eukaryotes, which can be utilized in downstream QIIME analyses just as any OTU table.
To build a tree utilizing the Silva 104 reference set, we will first create an alignment with the align_seqs.py module. The core Silva aligned set will be used as the template.
Use the following command with the rep_set.fna created in the OTU picking step above:
align_seqs.py -i rep_set.fna -t QIIME_files/core_aligned_set/core_Silva_aligned.fasta -o pynast_aligned/
Next, the alignment must be filtered. For 16S datasets, a Lanemask is usually applied to remove high entropy positions. QIIME has incorporated a dynamic entropy and gap calculation to the filter_alignment.py module, which removes the need for a Lanemask. To filter the alignment created above, use the following command:
filter_alignment.py -i pynast_aligned/rep_set_aligned.fna -o pynast_aligned/ -e 0.10 -g 0.80
In this case, the 10% most variable positions and positions that are greater than 80% gaps were removed (the -e and -g parameters respectively).
Finally, a tree can be built using make_phylogeny.py:
make_phylogeny.py -i pynast_aligned/rep_set_aligned_pfiltered.fasta -o rep_set.tre
Trees an OTU tables created can then be utilized in the downstream QIIME analyses as seen in the tutorial.
The Silva 104 reference set can be used in a workflow, such as pick_otus_through_otu_table.py. It is necessary to modify the qiime_parameters.txt file to correctly point to the Silva reference filepaths, and to use the dynamic alignment filtering rather than the 16S Lanemask. See the documentation for details about the qiime_parameters.txt file.
Parameters that should be modified:
- pick_otus:otu_picking_method uclust (should be set to uclust_ref if a reference based approach is desired)
- pick_otus:refseqs_fp (specify the filepath to the representative Silva 104 set, if reference based approach is desired)
- align_seqs:template_fp (specify the core aligned Silva 104 fasta file path)
- filter_alignment:lane_mask_fp (do not specify a lanemask filepath)
- filter_alignment:allowed_gap_frac 0.999999 (set to 0.80 instead of default)
- filter_alignment:entropy_threshold 0.10 (set to 0.10 if not already set)
- assign_taxonomy:id_to_taxonomy_fp (specify the taxonomy mapping file path, RDP version if RDP is the method of choice)
- assign_taxonomy:reference_seqs_fp (specify the Silva representative set file path)
These files have been modified from the Silva 104 release to help integration into the QIIME pipeline for marker gene (i.e. small ribosomal subunit) based analysis.
Versions of software used, apart from custom parsers:
uclust v1.2.22q version used for clustering Silva files. Primer Prospector (http://pprospector.sourceforge.net/) module clean_fasta.py was used to degas, remove spaces, and/or convert “U” to “T” in fasta files. fasttree 2.1.0 was used to construct the phylogenetic tree.
Core Silva aligned set generated by taking complete Silva 104 set, filtered to 80% identity with uclust, followed by filtering out positions that were greater than 99% gaps.
The representative set was generated by clustering the full Silva 104 release fasta file at 97% identity.
Taxonomy mapping files were generating by parsing taxonomy strings from the Silva fasta file. The RDP compatible file was created with a custom parser to get the required 6 levels of taxonomy, followed by hand curation to clean up empty levels of taxonomic definition.
The representative sequences were first filtered from the original Silva 104 alignment to remove positions that were > 90% gaps and entropy filtered to remove the 10% most entropic (variable) positions, and the resulting alignment was input to Fasttree to build the tree. Tree was then manually rooted between the Archaeal and Eukaryotic clades. Note that while this tree has performed reasonably well for phylogenetic analysis (i.e., Unifrac), the structure of the Eukaryotic domain of the tree of life is subject to ongoing debates and likely can not be resolved by the use of single gene markers, such as the SSU, alone.
