“Moving Pictures” tutorial - Multiple Interface Edition¶

Beta preview note¶

This guide is a beta preview of new functionality in QIIME 2 that aims to support development of multi-interface tutorials.

Introduction¶

In this tutorial you’ll use QIIME 2 to perform an analysis of human microbiome samples from two individuals at four body sites at five timepoints, the first of which immediately followed antibiotic usage. A study based on these samples was originally published in Caporaso et al. (2011). The data used in this tutorial were sequenced on an Illumina HiSeq using the Earth Microbiome Project hypervariable region 4 (V4) 16S rRNA sequencing protocol.

Sample metadata¶

Before starting the analysis, explore the sample metadata to familiarize yourself with the samples used in this study. The sample metadata is available as a Google Sheet. You can download this file as tab-separated text by selecting File > Download as > Tab-separated values. Alternatively, the following command will download the sample metadata as tab-separated text and save it in the file sample-metadata.tsv. This sample-metadata.tsv file is used throughout the rest of the tutorial.

from qiime2 import Metadata
from urllib import request

url = 'https://docs.qiime2.org/2021.11/data/tutorials/moving-pictures-usage/sample-metadata.tsv'
fn = 'sample-metadata.tsv'
request.urlretrieve(url, fn)
sample_metadata_md = Metadata.load(fn)

wget \
  -O 'sample-metadata.tsv' \
  'https://docs.qiime2.org/2021.11/data/tutorials/moving-pictures-usage/sample-metadata.tsv'

def md_factory():
    from urllib import request
    from qiime2 import Metadata
    fp, _ = request.urlretrieve(
        'https://data.qiime2.org/2024.2/tutorials/moving-pictures/sample_metadata.tsv',
    )

    return Metadata.load(fp)

sample_metadata = use.init_metadata('sample_metadata', md_factory)

Obtaining and importing data¶

Download the sequence reads that we’ll use in this analysis. In this tutorial we’ll work with a small subset of the complete sequence data so that the commands will run quickly.

import zipfile

url = 'https://docs.qiime2.org/2021.11/data/tutorials/moving-pictures-usage/emp-single-end-sequences.zip'
fn = 'emp-single-end-sequences.zip'
request.urlretrieve(url, fn)
with zipfile.ZipFile(fn) as zf:
    zf.extractall('emp-single-end-sequences')

wget \
  -O 'emp-single-end-sequences.zip' \
  'https://docs.qiime2.org/2021.11/data/tutorials/moving-pictures-usage/emp-single-end-sequences.zip'

unzip -d emp-single-end-sequences emp-single-end-sequences.zip

def emp_factory():
    import os
    import tempfile
    from urllib import request

    from q2_demux._format import EMPSingleEndDirFmt
    from q2_types.per_sample_sequences import FastqGzFormat

    base_url = 'https://data.qiime2.org/2024.2/tutorials/moving-pictures/'
    bc_url = base_url + 'emp-single-end-sequences/barcodes.fastq.gz'
    seqs_url = base_url + 'emp-single-end-sequences/sequences.fastq.gz'

    fmt = EMPSingleEndDirFmt(mode='w')

    with tempfile.TemporaryDirectory() as tmpdir:
        bc_fp = os.path.join(tmpdir, 'barcodes.fastq.gz')
        bc_fn, _ = request.urlretrieve(bc_url, bc_fp)

        seqs_fp = os.path.join(tmpdir, 'sequences.fastq.gz')
        seqs_fn, _ = request.urlretrieve(seqs_url, seqs_fp)

        fmt.barcodes.write_data(bc_fn, FastqGzFormat)
        fmt.sequences.write_data(seqs_fn, FastqGzFormat)

    fmt.validate()
    return fmt

raw_seqs = use.init_format('emp-single-end-sequences', emp_factory)

All data that is used as input to QIIME 2 is in form of QIIME 2 artifacts, which contain information about the type of data and the source of the data. So, the first thing we need to do is import these sequence data files into a QIIME 2 artifact.

The semantic type of this QIIME 2 artifact is EMPSingleEndSequences. EMPSingleEndSequences QIIME 2 artifacts contain sequences that are multiplexed, meaning that the sequences have not yet been assigned to samples (where the barcodes.fastq.gz contains the barcode read associated with each sequence in sequences.fastq.gz.) To learn about how to import sequence data in other formats, see the importing data tutorial.

from qiime2 import Artifact

emp_single_end_sequences = Artifact.import_data(
    'EMPSingleEndSequences',
    'emp-single-end-sequences',
)

qiime tools import \
  --type 'EMPSingleEndSequences' \
  --input-path emp-single-end-sequences \
  --output-path emp-single-end-sequences.qza

emp_single_end_sequences = use.import_from_format(
    'emp_single_end_sequences',
    'EMPSingleEndSequences',
    raw_seqs,
)

It is possible to check the UUID, type, and format of your newly-imported sequences, confirming that your import worked as expected:

print(emp_single_end_sequences.uuid)
print(emp_single_end_sequences.type)
print(emp_single_end_sequences.format)

qiime tools peek emp-single-end-sequences.qza

use.peek(emp_single_end_sequences)

Demultiplexing sequences¶

To demultiplex sequences we need to know which barcode sequence is associated with each sample. This information is contained in the sample metadata file. You can run the following commands to demultiplex the sequences (the demux emp-single command refers to the fact that these sequences are barcoded according to the Earth Microbiome Project protocol, and are single-end reads). The demux.qza QIIME 2 artifact will contain the demultiplexed sequences. The second output (demux-details.qza) presents Golay error correction details, and will not be explored in this tutorial (you can visualize these data using qiime metadata tabulate).

import qiime2.plugins.demux.actions as demux_actions

barcode_sequence_mdc = sample_metadata_md.get_column('barcode-sequence')
demux, demux_details = demux_actions.emp_single(
    seqs=emp_single_end_sequences,
    barcodes=barcode_sequence_mdc,
)

qiime demux emp-single \
  --i-seqs emp-single-end-sequences.qza \
  --m-barcodes-file sample-metadata.tsv \
  --m-barcodes-column barcode-sequence \
  --o-per-sample-sequences demux.qza \
  --o-error-correction-details demux-details.qza

barcode_sequence = use.get_metadata_column(
    'barcode_sequence', 'barcode-sequence', sample_metadata)

demux, demux_details = use.action(
    use.UsageAction(plugin_id='demux', action_id='emp_single'),
    use.UsageInputs(
        seqs=emp_single_end_sequences,
        barcodes=barcode_sequence,
    ),
    use.UsageOutputNames(
        per_sample_sequences='demux',
        error_correction_details='demux_details',
    ),
)

After demultiplexing, it’s useful to generate a summary of the demultiplexing results. This allows you to determine how many sequences were obtained per sample, and also to get a summary of the distribution of sequence qualities at each position in your sequence data.

demux_viz, = demux_actions.summarize(
    data=demux,
)

qiime demux summarize \
  --i-data demux.qza \
  --o-visualization demux.qzv

use.action(
    use.UsageAction(plugin_id='demux', action_id='summarize'),
    use.UsageInputs(data=demux),
    use.UsageOutputNames(visualization='demux'),
)

qiime tools view demux.qzv

Sequence quality control and feature table construction¶

QIIME 2 plugins are available for several quality control methods, including DADA2, Deblur, and basic quality-score-based filtering. In this tutorial we present this step using DADA2 and Deblur. These steps are interchangeable, so you can use whichever of these you prefer. The result of both of these methods will be a FeatureTable[Frequency] QIIME 2 artifact, which contains counts (frequencies) of each unique sequence in each sample in the dataset, and a FeatureData[Sequence] QIIME 2 artifact, which maps feature identifiers in the FeatureTable to the sequences they represent.

import qiime2.plugins.dada2.actions as dada2_actions

table, rep_seqs, stats = dada2_actions.denoise_single(
    demultiplexed_seqs=demux,
    trim_left=0,
    trunc_len=120,
)

qiime dada2 denoise-single \
  --i-demultiplexed-seqs demux.qza \
  --p-trim-left 0 \
  --p-trunc-len 120 \
  --o-representative-sequences rep-seqs.qza \
  --o-table table.qza \
  --o-denoising-stats stats.qza

rep_seqs_dada2, table_dada2, stats_dada2 = use.action(
    use.UsageAction(plugin_id='dada2', action_id='denoise_single'),
    use.UsageInputs(demultiplexed_seqs=demux, trim_left=0, trunc_len=120),
    use.UsageOutputNames(representative_sequences='rep_seqs',
                         table='table', denoising_stats='stats')
)

import qiime2.plugins.metadata.actions as metadata_actions

stats_dada2_md_md = stats.view(Metadata)
stats_viz, = metadata_actions.tabulate(
    input=stats_dada2_md_md,
)

qiime metadata tabulate \
  --m-input-file stats.qza \
  --o-visualization stats.qzv

stats_as_md = use.view_as_metadata('stats_dada2_md', stats_dada2)

use.action(
    use.UsageAction(plugin_id='metadata', action_id='tabulate'),
    use.UsageInputs(input=stats_as_md),
    use.UsageOutputNames(visualization='stats')
)

import qiime2.plugins.quality_filter.actions as quality_filter_actions

demux_filtered, demux_filter_stats = quality_filter_actions.q_score(
    demux=demux,
)

qiime quality-filter q-score \
  --i-demux demux.qza \
  --o-filtered-sequences demux-filtered.qza \
  --o-filter-stats demux-filter-stats.qza

filtered_seqs, filter_stats = use.action(
    use.UsageAction(plugin_id='quality_filter', action_id='q_score'),
    use.UsageInputs(demux=demux),
    use.UsageOutputNames(filtered_sequences='demux_filtered',
                         filter_stats='demux_filter_stats')
)

import qiime2.plugins.deblur.actions as deblur_actions

table_deblur, rep_seqs_deblur, deblur_stats = deblur_actions.denoise_16S(
    demultiplexed_seqs=demux_filtered,
    trim_length=120,
    sample_stats=True,
)

qiime deblur denoise-16S \
  --i-demultiplexed-seqs demux-filtered.qza \
  --p-trim-length 120 \
  --p-sample-stats \
  --o-representative-sequences rep-seqs-deblur.qza \
  --o-table table-deblur.qza \
  --o-stats deblur-stats.qza

rep_seqs_deblur, table_deblur, stats_deblur = use.action(
     use.UsageAction(plugin_id='deblur', action_id='denoise_16S'),
     use.UsageInputs(demultiplexed_seqs=filtered_seqs, trim_length=120,
                     sample_stats=True),
     use.UsageOutputNames(representative_sequences='rep_seqs_deblur',
                          table='table_deblur', stats='deblur_stats'),
)

filter_stats_md = demux_filter_stats.view(Metadata)
demux_filter_stats_viz, = metadata_actions.tabulate(
    input=filter_stats_md,
)
deblur_stats_viz, = deblur_actions.visualize_stats(
    deblur_stats=deblur_stats,
)

qiime metadata tabulate \
  --m-input-file demux-filter-stats.qza \
  --o-visualization demux-filter-stats.qzv
qiime deblur visualize-stats \
  --i-deblur-stats deblur-stats.qza \
  --o-visualization deblur-stats.qzv

filter_stats_as_md = use.view_as_metadata('filter_stats', filter_stats)

use.action(
     use.UsageAction(plugin_id='metadata', action_id='tabulate'),
     use.UsageInputs(input=filter_stats_as_md),
     use.UsageOutputNames(visualization='demux_filter_stats'),
)

use.action(
     use.UsageAction(plugin_id='deblur', action_id='visualize_stats'),
     use.UsageInputs(deblur_stats=stats_deblur),
     use.UsageOutputNames(visualization='deblur_stats'),
)

# q2cli:
# mv rep-seqs-deblur.qza rep-seqs.qza
# mv table-deblur.qza table.qza
# 
# Artifact API:
# table = table_deblur
# rep_seqs = rep_seqs_deblur

# q2cli:
# mv rep-seqs-deblur.qza rep-seqs.qza
# mv table-deblur.qza table.qza
# Artifact API:
# table = table_deblur
# rep_seqs = rep_seqs_deblur

use.comment('q2cli:')
use.comment('mv rep-seqs-deblur.qza rep-seqs.qza')
use.comment('mv table-deblur.qza table.qza')
use.comment('')
use.comment('Artifact API:')
use.comment('table = table_deblur')
use.comment('rep_seqs = rep_seqs_deblur')

FeatureTable and FeatureData summaries¶

After the quality filtering step completes, you’ll want to explore the resulting data. You can do this using the following two commands, which will create visual summaries of the data. The feature-table summarize command will give you information on how many sequences are associated with each sample and with each feature, histograms of those distributions, and some related summary statistics. The feature-table tabulate-seqs command will provide a mapping of feature IDs to sequences, and provide links to easily BLAST each sequence against the NCBI nt database. The latter visualization will be very useful later in the tutorial, when you want to learn more about specific features that are important in the data set.

import qiime2.plugins.feature_table.actions as feature_table_actions

table_viz, = feature_table_actions.summarize(
    table=table,
    sample_metadata=sample_metadata_md,
)
rep_seqs_viz, = feature_table_actions.tabulate_seqs(
    data=rep_seqs,
)

qiime feature-table summarize \
  --i-table table.qza \
  --m-sample-metadata-file sample-metadata.tsv \
  --o-visualization table.qzv
qiime feature-table tabulate-seqs \
  --i-data rep-seqs.qza \
  --o-visualization rep-seqs.qzv

use.action(
     use.UsageAction(plugin_id='feature_table', action_id='summarize'),
     use.UsageInputs(table=table_dada2, sample_metadata=sample_metadata),
     use.UsageOutputNames(visualization='table'),
)

use.action(
     use.UsageAction(plugin_id='feature_table', action_id='tabulate_seqs'),
     use.UsageInputs(data=rep_seqs_dada2),
     use.UsageOutputNames(visualization='rep_seqs'),
)

Generate a tree for phylogenetic diversity analyses¶

QIIME supports several phylogenetic diversity metrics, including Faith’s Phylogenetic Diversity and weighted and unweighted UniFrac. In addition to counts of features per sample (i.e., the data in the FeatureTable[Frequency] QIIME 2 artifact), these metrics require a rooted phylogenetic tree relating the features to one another. This information will be stored in a Phylogeny[Rooted] QIIME 2 artifact. To generate a phylogenetic tree we will use align-to-tree-mafft-fasttree pipeline from the q2-phylogeny plugin.

First, the pipeline uses the mafft program to perform a multiple sequence alignment of the sequences in our FeatureData[Sequence] to create a FeatureData[AlignedSequence] QIIME 2 artifact. Next, the pipeline masks (or filters) the alignment to remove positions that are highly variable. These positions are generally considered to add noise to a resulting phylogenetic tree. Following that, the pipeline applies FastTree to generate a phylogenetic tree from the masked alignment. The FastTree program creates an unrooted tree, so in the final step in this section midpoint rooting is applied to place the root of the tree at the midpoint of the longest tip-to-tip distance in the unrooted tree.

import qiime2.plugins.phylogeny.actions as phylogeny_actions

action_results = phylogeny_actions.align_to_tree_mafft_fasttree(
    sequences=rep_seqs,
)
aligned_rep_seqs = action_results.alignment
masked_aligned_rep_seqs = action_results.masked_alignment
unrooted_tree = action_results.tree
rooted_tree = action_results.rooted_tree

qiime phylogeny align-to-tree-mafft-fasttree \
  --i-sequences rep-seqs.qza \
  --output-dir phylogeny-align-to-tree-mafft-fasttree

_, _, _, rooted_tree = use.action(
     use.UsageAction(plugin_id='phylogeny', action_id='align_to_tree_mafft_fasttree'),
     use.UsageInputs(sequences=rep_seqs_dada2),
     use.UsageOutputNames(alignment='aligned_rep_seqs',
                          masked_alignment='masked_aligned_rep_seqs',
                          tree='unrooted_tree', rooted_tree='rooted_tree'),
)

Alpha and beta diversity analysis¶

QIIME 2’s diversity analyses are available through the q2-diversity plugin, which supports computing alpha and beta diversity metrics, applying related statistical tests, and generating interactive visualizations. We’ll first apply the core-metrics-phylogenetic method, which rarefies a FeatureTable[Frequency] to a user-specified depth, computes several alpha and beta diversity metrics, and generates principle coordinates analysis (PCoA) plots using Emperor for each of the beta diversity metrics. The metrics computed by default are:

• Alpha diversity

  • Shannon’s diversity index (a quantitative measure of community richness)

  • Observed Features (a qualitative measure of community richness)

  • Faith’s Phylogenetic Diversity (a qualitative measure of community richness that incorporates phylogenetic relationships between the features)

  • Evenness (or Pielou’s Evenness; a measure of community evenness)

• Beta diversity

  • Jaccard distance (a qualitative measure of community dissimilarity)

  • Bray-Curtis distance (a quantitative measure of community dissimilarity)

  • unweighted UniFrac distance (a qualitative measure of community dissimilarity that incorporates phylogenetic relationships between the features)

  • weighted UniFrac distance (a quantitative measure of community dissimilarity that incorporates phylogenetic relationships between the features)

An important parameter that needs to be provided to this script is --p-sampling-depth, which is the even sampling (i.e. rarefaction) depth. Because most diversity metrics are sensitive to different sampling depths across different samples, this script will randomly subsample the counts from each sample to the value provided for this parameter. For example, if you provide --p-sampling-depth 500, this step will subsample the counts in each sample without replacement so that each sample in the resulting table has a total count of 500. If the total count for any sample(s) are smaller than this value, those samples will be dropped from the diversity analysis. Choosing this value is tricky. We recommend making your choice by reviewing the information presented in the table.qzv file that was created above. Choose a value that is as high as possible (so you retain more sequences per sample) while excluding as few samples as possible.

import qiime2.plugins.diversity.actions as diversity_actions

action_results = diversity_actions.core_metrics_phylogenetic(
    phylogeny=rooted_tree,
    table=table,
    sampling_depth=1103,
    metadata=sample_metadata_md,
)
rarefied_table = action_results.rarefied_table
faith_pd_vector = action_results.faith_pd_vector
observed_features_vector = action_results.observed_features_vector
shannon_vector = action_results.shannon_vector
evenness_vector = action_results.evenness_vector
unweighted_unifrac_distance_matrix = action_results.unweighted_unifrac_distance_matrix
weighted_unifrac_distance_matrix = action_results.weighted_unifrac_distance_matrix
jaccard_distance_matrix = action_results.jaccard_distance_matrix
bray_curtis_distance_matrix = action_results.bray_curtis_distance_matrix
unweighted_unifrac_pcoa_results = action_results.unweighted_unifrac_pcoa_results
weighted_unifrac_pcoa_results = action_results.weighted_unifrac_pcoa_results
jaccard_pcoa_results = action_results.jaccard_pcoa_results
bray_curtis_pcoa_results = action_results.bray_curtis_pcoa_results
unweighted_unifrac_emperor_viz = action_results.unweighted_unifrac_emperor
weighted_unifrac_emperor_viz = action_results.weighted_unifrac_emperor
jaccard_emperor_viz = action_results.jaccard_emperor
bray_curtis_emperor_viz = action_results.bray_curtis_emperor

qiime diversity core-metrics-phylogenetic \
  --i-phylogeny phylogeny-align-to-tree-mafft-fasttree/rooted_tree.qza \
  --i-table table.qza \
  --p-sampling-depth 1103 \
  --m-metadata-file sample-metadata.tsv \
  --output-dir diversity-core-metrics-phylogenetic

core_metrics_results = use.action(
     use.UsageAction(plugin_id='diversity', action_id='core_metrics_phylogenetic'),
     use.UsageInputs(phylogeny=rooted_tree, table=table_dada2,
                     sampling_depth=1103, metadata=sample_metadata),
     use.UsageOutputNames(rarefied_table='rarefied_table',
                          faith_pd_vector='faith_pd_vector',
                          observed_features_vector='observed_features_vector',
                          shannon_vector='shannon_vector',
                          evenness_vector='evenness_vector',
                          unweighted_unifrac_distance_matrix='unweighted_unifrac_distance_matrix',
                          weighted_unifrac_distance_matrix='weighted_unifrac_distance_matrix',
                          jaccard_distance_matrix='jaccard_distance_matrix',
                          bray_curtis_distance_matrix='bray_curtis_distance_matrix',
                          unweighted_unifrac_pcoa_results='unweighted_unifrac_pcoa_results',
                          weighted_unifrac_pcoa_results='weighted_unifrac_pcoa_results',
                          jaccard_pcoa_results='jaccard_pcoa_results',
                          bray_curtis_pcoa_results='bray_curtis_pcoa_results',
                          unweighted_unifrac_emperor='unweighted_unifrac_emperor',
                          weighted_unifrac_emperor='weighted_unifrac_emperor',
                          jaccard_emperor='jaccard_emperor',
                          bray_curtis_emperor='bray_curtis_emperor'),
)
faith_pd_vec = core_metrics_results.faith_pd_vector
evenness_vec = core_metrics_results.evenness_vector
unweighted_unifrac_dm = core_metrics_results.unweighted_unifrac_distance_matrix
unweighted_unifrac_pcoa = core_metrics_results.unweighted_unifrac_pcoa_results
bray_curtis_pcoa=core_metrics_results.bray_curtis_pcoa_results

Here we set the --p-sampling-depth parameter to 1103. This value was chosen based on the number of sequences in the L3S313 sample because it’s close to the number of sequences in the next few samples that have higher sequence counts, and because it is considerably higher (relatively) than the number of sequences in the samples that have fewer sequences. This will allow us to retain most of our samples. The three samples that have fewer sequences will be dropped from the core-metrics-phylogenetic analyses and anything that uses these results. It is worth noting that all three of these samples are “right palm” samples. Losing a disproportionate number of samples from one metadata category is not ideal. However, we are dropping a small enough number of samples here that this felt like the best compromise between total sequences analyzed and number of samples retained.

After computing diversity metrics, we can begin to explore the microbial composition of the samples in the context of the sample metadata. This information is present in the sample metadata file that was downloaded earlier.

We’ll first test for associations between categorical metadata columns and alpha diversity data. We’ll do that here for the Faith Phylogenetic Diversity (a measure of community richness) and evenness metrics.

faith_pd_group_significance_viz, = diversity_actions.alpha_group_significance(
    alpha_diversity=faith_pd_vector,
    metadata=sample_metadata_md,
)
evenness_group_significance_viz, = diversity_actions.alpha_group_significance(
    alpha_diversity=evenness_vector,
    metadata=sample_metadata_md,
)

qiime diversity alpha-group-significance \
  --i-alpha-diversity diversity-core-metrics-phylogenetic/faith_pd_vector.qza \
  --m-metadata-file sample-metadata.tsv \
  --o-visualization faith-pd-group-significance.qzv
qiime diversity alpha-group-significance \
  --i-alpha-diversity diversity-core-metrics-phylogenetic/evenness_vector.qza \
  --m-metadata-file sample-metadata.tsv \
  --o-visualization evenness-group-significance.qzv

use.action(
     use.UsageAction(plugin_id='diversity', action_id='alpha_group_significance'),
     use.UsageInputs(alpha_diversity=faith_pd_vec, metadata=sample_metadata),
     use.UsageOutputNames(visualization='faith_pd_group_significance'),
)

use.action(
     use.UsageAction(plugin_id='diversity', action_id='alpha_group_significance'),
     use.UsageInputs(alpha_diversity=evenness_vec, metadata=sample_metadata),
     use.UsageOutputNames(visualization='evenness_group_significance'),
)

In this data set, no continuous sample metadata columns (e.g., days-since-experiment-start) are correlated with alpha diversity, so we won’t test for those associations here. If you’re interested in performing those tests (for this data set, or for others), you can use the qiime diversity alpha-correlation command.

Next we’ll analyze sample composition in the context of categorical metadata using PERMANOVA (first described in Anderson (2001)) using the beta-group-significance command. The following commands will test whether distances between samples within a group, such as samples from the same body site (e.g., gut), are more similar to each other then they are to samples from the other groups (e.g., tongue, left palm, and right palm). If you call this command with the --p-pairwise parameter, as we’ll do here, it will also perform pairwise tests that will allow you to determine which specific pairs of groups (e.g., tongue and gut) differ from one another, if any. This command can be slow to run, especially when passing --p-pairwise, since it is based on permutation tests. So, unlike the previous commands, we’ll run beta-group-significance on specific columns of metadata that we’re interested in exploring, rather than all metadata columns to which it is applicable. Here we’ll apply this to our unweighted UniFrac distances, using two sample metadata columns, as follows.

body_site_mdc = sample_metadata_md.get_column('body-site')
unweighted_unifrac_body_site_group_significance_viz, = diversity_actions.beta_group_significance(
    distance_matrix=unweighted_unifrac_distance_matrix,
    metadata=body_site_mdc,
    pairwise=True,
)
subject_mdc = sample_metadata_md.get_column('subject')
unweighted_unifrac_subject_group_significance_viz, = diversity_actions.beta_group_significance(
    distance_matrix=unweighted_unifrac_distance_matrix,
    metadata=subject_mdc,
    pairwise=True,
)

qiime diversity beta-group-significance \
  --i-distance-matrix diversity-core-metrics-phylogenetic/unweighted_unifrac_distance_matrix.qza \
  --m-metadata-file sample-metadata.tsv \
  --m-metadata-column body-site \
  --p-pairwise \
  --o-visualization unweighted-unifrac-body-site-group-significance.qzv
qiime diversity beta-group-significance \
  --i-distance-matrix diversity-core-metrics-phylogenetic/unweighted_unifrac_distance_matrix.qza \
  --m-metadata-file sample-metadata.tsv \
  --m-metadata-column subject \
  --p-pairwise \
  --o-visualization unweighted-unifrac-subject-group-significance.qzv

body_site_col = use.get_metadata_column('body_site', 'body-site', sample_metadata)

use.action(
     use.UsageAction(plugin_id='diversity', action_id='beta_group_significance'),
     use.UsageInputs(distance_matrix=unweighted_unifrac_dm,
                     metadata=body_site_col, pairwise=True),
     use.UsageOutputNames(visualization='unweighted_unifrac_body_site_group_significance'),
)

subject_col = use.get_metadata_column('subject', 'subject', sample_metadata)

use.action(
     use.UsageAction(plugin_id='diversity', action_id='beta_group_significance'),
     use.UsageInputs(distance_matrix=unweighted_unifrac_dm,
                     metadata=subject_col, pairwise=True),
     use.UsageOutputNames(visualization='unweighted_unifrac_subject_group_significance'),
)

Again, none of the continuous sample metadata that we have for this data set are correlated with sample composition, so we won’t test for those associations here. If you’re interested in performing those tests, you can use the qiime metadata distance-matrix in combination with qiime diversity mantel and qiime diversity bioenv commands.

Finally, ordination is a popular approach for exploring microbial community composition in the context of sample metadata. We can use the Emperor tool to explore principal coordinates (PCoA) plots in the context of sample metadata. While our core-metrics-phylogenetic command did already generate some Emperor plots, we want to pass an optional parameter, --p-custom-axes, which is very useful for exploring time series data. The PCoA results that were used in core-metrics-phylogeny are also available, making it easy to generate new visualizations with Emperor. We will generate Emperor plots for unweighted UniFrac and Bray-Curtis so that the resulting plot will contain axes for principal coordinate 1, principal coordinate 2, and days since the experiment start. We will use that last axis to explore how these samples changed over time.

import qiime2.plugins.emperor.actions as emperor_actions

unweighted_unifrac_emperor_days_since_experiment_start_viz, = emperor_actions.plot(
    pcoa=unweighted_unifrac_pcoa_results,
    metadata=sample_metadata_md,
    custom_axes=['days-since-experiment-start'],
)
bray_curtis_emperor_days_since_experiment_start_viz, = emperor_actions.plot(
    pcoa=bray_curtis_pcoa_results,
    metadata=sample_metadata_md,
    custom_axes=['days-since-experiment-start'],
)

qiime emperor plot \
  --i-pcoa diversity-core-metrics-phylogenetic/unweighted_unifrac_pcoa_results.qza \
  --m-metadata-file sample-metadata.tsv \
  --p-custom-axes days-since-experiment-start \
  --o-visualization unweighted-unifrac-emperor-days-since-experiment-start.qzv
qiime emperor plot \
  --i-pcoa diversity-core-metrics-phylogenetic/bray_curtis_pcoa_results.qza \
  --m-metadata-file sample-metadata.tsv \
  --p-custom-axes days-since-experiment-start \
  --o-visualization bray-curtis-emperor-days-since-experiment-start.qzv

use.action(
     use.UsageAction(plugin_id='emperor', action_id='plot'),
     use.UsageInputs(pcoa=unweighted_unifrac_pcoa, metadata=sample_metadata,
                     custom_axes=['days-since-experiment-start']),
     use.UsageOutputNames(visualization='unweighted-unifrac-emperor-days-since-experiment-start'),
)

use.action(
     use.UsageAction(plugin_id='emperor', action_id='plot'),
     use.UsageInputs(pcoa=bray_curtis_pcoa, metadata=sample_metadata,
                     custom_axes=['days-since-experiment-start']),
     use.UsageOutputNames(visualization='bray-curtis-emperor-days-since-experiment-start'),
)

Alpha rarefaction plotting¶

In this section we’ll explore alpha diversity as a function of sampling depth using the qiime diversity alpha-rarefaction visualizer. This visualizer computes one or more alpha diversity metrics at multiple sampling depths, in steps between 1 (optionally controlled with --p-min-depth) and the value provided as --p-max-depth. At each sampling depth step, 10 rarefied tables will be generated, and the diversity metrics will be computed for all samples in the tables. The number of iterations (rarefied tables computed at each sampling depth) can be controlled with --p-iterations. Average diversity values will be plotted for each sample at each even sampling depth, and samples can be grouped based on metadata in the resulting visualization if sample metadata is provided with the --m-metadata-file parameter.

alpha_rarefaction_viz, = diversity_actions.alpha_rarefaction(
    table=table,
    phylogeny=rooted_tree,
    max_depth=4000,
    metadata=sample_metadata_md,
)

qiime diversity alpha-rarefaction \
  --i-table table.qza \
  --i-phylogeny phylogeny-align-to-tree-mafft-fasttree/rooted_tree.qza \
  --p-max-depth 4000 \
  --m-metadata-file sample-metadata.tsv \
  --o-visualization alpha-rarefaction.qzv

use.action(
    use.UsageAction(plugin_id='diversity', action_id='alpha_rarefaction'),
    use.UsageInputs(table=table_dada2, phylogeny=rooted_tree,
                    max_depth=4000, metadata=sample_metadata),
    use.UsageOutputNames(visualization='alpha_rarefaction'),
)

The visualization will have two plots. The top plot is an alpha rarefaction plot, and is primarily used to determine if the richness of the samples has been fully observed or sequenced. If the lines in the plot appear to “level out” (i.e., approach a slope of zero) at some sampling depth along the x-axis, that suggests that collecting additional sequences beyond that sampling depth would not be likely to result in the observation of additional features. If the lines in a plot don’t level out, this may be because the richness of the samples hasn’t been fully observed yet (because too few sequences were collected), or it could be an indicator that a lot of sequencing error remains in the data (which is being mistaken for novel diversity).

The bottom plot in this visualization is important when grouping samples by metadata. It illustrates the number of samples that remain in each group when the feature table is rarefied to each sampling depth. If a given sampling depth d is larger than the total frequency of a sample s (i.e., the number of sequences that were obtained for sample s), it is not possible to compute the diversity metric for sample s at sampling depth d. If many of the samples in a group have lower total frequencies than d, the average diversity presented for that group at d in the top plot will be unreliable because it will have been computed on relatively few samples. When grouping samples by metadata, it is therefore essential to look at the bottom plot to ensure that the data presented in the top plot is reliable.

Taxonomic analysis¶

In the next sections we’ll begin to explore the taxonomic composition of the samples, and again relate that to sample metadata. The first step in this process is to assign taxonomy to the sequences in our FeatureData[Sequence] QIIME 2 artifact. We’ll do that using a pre-trained Naive Bayes classifier and the q2-feature-classifier plugin. This classifier was trained on the Greengenes 13_8 99% OTUs, where the sequences have been trimmed to only include 250 bases from the region of the 16S that was sequenced in this analysis (the V4 region, bound by the 515F/806R primer pair). We’ll apply this classifier to our sequences, and we can generate a visualization of the resulting mapping from sequence to taxonomy.

url = 'https://docs.qiime2.org/2021.11/data/tutorials/moving-pictures-usage/gg-13-8-99-515-806-nb-classifier.qza'
fn = 'gg-13-8-99-515-806-nb-classifier.qza'
request.urlretrieve(url, fn)
gg_13_8_99_515_806_nb_classifier = Artifact.load(fn)

wget \
  -O 'gg-13-8-99-515-806-nb-classifier.qza' \
  'https://docs.qiime2.org/2021.11/data/tutorials/moving-pictures-usage/gg-13-8-99-515-806-nb-classifier.qza'

def classifier_factory():
    from urllib import request
    from qiime2 import Artifact
    fp, _ = request.urlretrieve(
        'https://data.qiime2.org/2024.2/common/gg-13-8-99-515-806-nb-classifier.qza',
    )

    return Artifact.load(fp)

classifier = use.init_artifact('gg-13-8-99-515-806-nb-classifier', classifier_factory)

import qiime2.plugins.feature_classifier.actions as feature_classifier_actions

taxonomy, = feature_classifier_actions.classify_sklearn(
    classifier=gg_13_8_99_515_806_nb_classifier,
    reads=rep_seqs,
)
taxonomy_as_md_md = taxonomy.view(Metadata)
taxonomy_viz, = metadata_actions.tabulate(
    input=taxonomy_as_md_md,
)

qiime feature-classifier classify-sklearn \
  --i-classifier gg-13-8-99-515-806-nb-classifier.qza \
  --i-reads rep-seqs.qza \
  --o-classification taxonomy.qza
qiime metadata tabulate \
  --m-input-file taxonomy.qza \
  --o-visualization taxonomy.qzv

taxonomy, = use.action(
     use.UsageAction(plugin_id='feature_classifier', action_id='classify_sklearn'),
     use.UsageInputs(classifier=classifier, reads=rep_seqs_dada2),
     use.UsageOutputNames(classification='taxonomy'),
)

taxonomy_as_md = use.view_as_metadata('taxonomy_as_md', taxonomy)

use.action(
     use.UsageAction(plugin_id='metadata', action_id='tabulate'),
     use.UsageInputs(input=taxonomy_as_md),
     use.UsageOutputNames(visualization='taxonomy'),
)

Next, we can view the taxonomic composition of our samples with interactive bar plots. Generate those plots with the following command and then open the visualization.

import qiime2.plugins.taxa.actions as taxa_actions

taxa_bar_plots_viz, = taxa_actions.barplot(
    table=table,
    taxonomy=taxonomy,
    metadata=sample_metadata_md,
)

qiime taxa barplot \
  --i-table table.qza \
  --i-taxonomy taxonomy.qza \
  --m-metadata-file sample-metadata.tsv \
  --o-visualization taxa-bar-plots.qzv

use.action(
     use.UsageAction(plugin_id='taxa', action_id='barplot'),
     use.UsageInputs(table=table_dada2, taxonomy=taxonomy,
                      metadata=sample_metadata),
     use.UsageOutputNames(visualization='taxa_bar_plots'),
)

Differential abundance testing with ANCOM-BC¶

ANCOM-BC can be applied to identify features that are differentially abundant (i.e. present in different abundances) across sample groups. As with any bioinformatics method, you should be aware of the assumptions and limitations of ANCOM-BC before using it. We recommend reviewing the ANCOM-BC paper before using this method.

ANCOM-BC is a compositionally-aware linear regression model that allows for testing differentially abundant features across groups while also implementing bias correction, and is currently implemented in the q2-composition plugin.

Because we expect a lot of features to change in abundance across body sites, in this tutorial we’ll filter our full feature table to only contain gut samples. We’ll then apply ANCOM-BC to determine which, if any, sequence variants and genera are differentially abundant across the gut samples of our two subjects.

We’ll start by creating a feature table that contains only the gut samples. (To learn more about filtering, see the Filtering Data tutorial.)

gut_table, = feature_table_actions.filter_samples(
    table=table,
    metadata=sample_metadata_md,
    where='[body-site]="gut"',
)

qiime feature-table filter-samples \
  --i-table table.qza \
  --m-metadata-file sample-metadata.tsv \
  --p-where '[body-site]="gut"' \
  --o-filtered-table gut-table.qza

gut_table, = use.action(
     use.UsageAction(plugin_id='feature_table', action_id='filter_samples'),
     use.UsageInputs(table=table_dada2, metadata=sample_metadata,
                     where='[body-site]="gut"'),
     use.UsageOutputNames(filtered_table='gut_table'),
)

ANCOM-BC operates on a FeatureTable[Frequency] QIIME 2 artifact. We can run ANCOM-BC on the subject column to determine what features differ in abundance across gut samples of the two subjects.

import qiime2.plugins.composition.actions as composition_actions

ancombc_subject, = composition_actions.ancombc(
    table=gut_table,
    metadata=sample_metadata_md,
    formula='subject',
)
da_barplot_subject_viz, = composition_actions.da_barplot(
    data=ancombc_subject,
    significance_threshold=0.001,
)

qiime composition ancombc \
  --i-table gut-table.qza \
  --m-metadata-file sample-metadata.tsv \
  --p-formula subject \
  --o-differentials ancombc-subject.qza
qiime composition da-barplot \
  --i-data ancombc-subject.qza \
  --p-significance-threshold 0.001 \
  --o-visualization da-barplot-subject.qzv

ancombc_subject, = use.action(
     use.UsageAction(plugin_id='composition', action_id='ancombc'),
     use.UsageInputs(table=gut_table, metadata=sample_metadata, formula='subject'),
     use.UsageOutputNames(differentials='ancombc_subject'),
)

use.action(
   use.UsageAction(plugin_id='composition', action_id='da_barplot'),
   use.UsageInputs(data=ancombc_subject, significance_threshold=0.001),
   use.UsageOutputNames(visualization='da_barplot_subject'),
)

We’re also often interested in performing a differential abundance test at a specific taxonomic level. To do this, we can collapse the features in our FeatureTable[Frequency] at the taxonomic level of interest, and then re-run the above steps. In this tutorial, we collapse our feature table at the genus level (i.e. level 6 of the Greengenes taxonomy).

gut_table_l6, = taxa_actions.collapse(
    table=gut_table,
    taxonomy=taxonomy,
    level=6,
)
l6_ancombc_subject, = composition_actions.ancombc(
    table=gut_table_l6,
    metadata=sample_metadata_md,
    formula='subject',
)
l6_da_barplot_subject_viz, = composition_actions.da_barplot(
    data=l6_ancombc_subject,
    significance_threshold=0.001,
)

qiime taxa collapse \
  --i-table gut-table.qza \
  --i-taxonomy taxonomy.qza \
  --p-level 6 \
  --o-collapsed-table gut-table-l6.qza
qiime composition ancombc \
  --i-table gut-table-l6.qza \
  --m-metadata-file sample-metadata.tsv \
  --p-formula subject \
  --o-differentials l6-ancombc-subject.qza
qiime composition da-barplot \
  --i-data l6-ancombc-subject.qza \
  --p-significance-threshold 0.001 \
  --o-visualization l6-da-barplot-subject.qzv

l6_gut_table, = use.action(
     use.UsageAction(plugin_id='taxa', action_id='collapse'),
     use.UsageInputs(table=gut_table, taxonomy=taxonomy, level=6),
     use.UsageOutputNames(collapsed_table='gut_table_l6'),
)

l6_ancombc_subject, = use.action(
     use.UsageAction(plugin_id='composition', action_id='ancombc'),
     use.UsageInputs(table=l6_gut_table, metadata=sample_metadata, formula='subject'),
     use.UsageOutputNames(differentials='l6_ancombc_subject'),
)

use.action(
     use.UsageAction(plugin_id='composition', action_id='da_barplot'),
     use.UsageInputs(data=l6_ancombc_subject, significance_threshold=0.001),
     use.UsageOutputNames(visualization='l6_da_barplot_subject'),
)