Basic Microbiome Analysis with QIIME

34
Basic Microbiome Analysis with QIIME Patricio Jeraldo and Bryan White Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

description

Basic Microbiome Analysis with QIIME. Patricio Jeraldo and Bryan White. Exercise. In this exercise we will do the following: Calculate sample diversity ( - diversity), and test if different sample types have different numbers of OTUs (species). - PowerPoint PPT Presentation

Transcript of Basic Microbiome Analysis with QIIME

Page 1: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME

Patricio Jeraldo and Bryan White

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Page 2: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Exercise

In this exercise we will do the following:

1. Calculate sample diversity (-diversity), and test if different sample types have different numbers of OTUs (species).

2. Calculate differences in microbial community structure (-diversity); in particular, we will compare OTU composition and abundance between samples and sample types.

3. Compute statistical support for observed differences between sample types.

4. Plot taxonomy composition across samples.

5. Test for potential microbial markers.

Page 3: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 0A: Accessing the IGB BioclusterOpen Putty.exe

In the hostname textbox type:

biocluster.igb.Illinois.edu

Click Open

If popup appears, Click Yes

Enter login credentials assigned to you; example, user class45.

Now you are all set!

Page 4: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 0B: Lab SetupThe lab is located in the following directory:

~/mayo/white

This directory contains the finished version of the lab (i.e. the version of the lab after the tutorial). Consult it if you unsure about your runs. You don’t have write permissions to the lab directory. Create a working directory of this lab in your home directory for your output to be stored. Note ~ is a symbol in unix paths referring to your home directory. Copy the files

Make sure you login to a machine on the cluster using the qsub command. The exact syntax for this command is given below. This particular command will login you into a computer with 4 cpus with an interactive session. You only need to do this once.

Page 5: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 0C: Lab Setup$ qsub –I –l ncpus=4 # Login to a computer on cluster.

$ mkdir –p ~/white/results

# Make results directory in our working directory.

# –p indicates to create ~/white if it doesn’t exist.

$ cp ~/mayo/white/data/* ~/white

# Copy data to your working directory.

$ cd ~/white # Change directory to our working directory.

$ module load qiime # We will need QIIME for this lab.

Page 6: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 0D: Shared Desktop Directory

For viewing and manipulating files on the classroom computers, we provide a shared directory in the following folder on the desktop:

classes/mayo

In today’s lab, we will be using the following folder in the shared directory:

classes/mayo/white

Page 7: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Interstitial CystitisInterstitial cystitis (IC) is a chronic inflammation of the bowels. In this exercise, we will examine differences between the microbiota of women with and without IC to understand the effect IC has on the community.

Our data consists of 16S sequencing of stools samples from 8 women with IC and 7 without it. Using QIMME, we will examine

Using this data, we will test the hypothesis that IC induces significant change in gut microbiota. Additionally, we will examine whether or not there is a change in the community and what bacteria are implicated in causing such change.

Page 8: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 1A: Dataset CharacteristicsICF.biomThe ICF.biom file is an OTU observation file.

It is a matrix of observed OTUs, or species, for each sample, annotated with their taxonomy.

The ICF.biom file was created using our own TORNADO pipeline for 16S reads: quality check, chimera check, align, assign taxonomy and cluster to 97% similarity to find OTUs

The TORNADO pipeline can take from HOURS to DAYS depending on the complexity of the project.

Page 9: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 1B: Dataset Characteristics ICF.mapping.txtThe mapping file contains metadata associated with samples.

Let us examine the file using the Unix cat command.$ cat ICF.mapping.txt # print file contents to screen

#SampleID Barcode Dx SubjectID DescriptionICF-1 GGATCGCAGATC Control 1 IC_fecal1ICF-2 GCTGATGAGCTG Control 2 IC_fecal2ICF-3 AGCTGTTGTTTG Control 3 IC_fecal3ICF-4 GGATGGTGTTGC IC 4 IC_fecal4

The most important column to us.

Output:

Page 10: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 1C: Dataset Characteristics ICF.treeThe ICF.tree file is a Newick-formatted phylogenetic tree file.

It contains phylogenetic relationships between the OTUs found in our samples.

It is another output of the 16S pipeline required for various comparison metrics.

Page 11: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 1D: Dataset Characteristics params.txtThe params.txt file contains alternative parameters to run QIIME.

Let us examine the file using the Unix cat command.$ cat params.txt# print file contents to screen

beta_diversity:metrics bray_curtis,unweighted_unifrac,weighted_unifracalpha_diversity:metrics chao1,goods_coverage,observed_species,shannon,simpson,PD_whole_tree

Output:

Page 12: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 2: Get Basic StatisticsThe first step we will do is to get some basic statistics on our ICF.biom file.

We will use the per-library_stats.py script in QIIME to do this.$ per_library_stats.py –i ICF.biom # Get stats.

Num samples: 15Num otus: 260Num observations (sequences): 399985.0Table density (fraction of non-zero values): 0.6082Seqs/sample summary: Min: 10267.0 Max: 48123.0

Output:

Page 13: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 3: Calculating DiversityFor this next step, let us measure the diversity of the samples.

We will use the number from the previous slide so that, for comparison purposes, all samples will have the same number of sequences.

We will use the alpha_rarefaction.py script in QIIME to do this.

Results are located in ~/white/results/alpha_diversity.

$ alpha_rarefaction.py –i ICF.biom –t ICF.tree –m ICF.mapping.txt –o results/alpha_diversity –p params.txt –e 10267

This calculation will take from 5 - 7 min to complete.

Page 14: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 4: Calculating DiversityFor this next step, let us compare samples using their composition.

We will specify we are interested in the Dx column.

We will use the beta_diversity_through_plots.py script in QIIME to do this.

Results are located in ~/white/results/beta_diversity. We will use these results later in the tutorial.

$ beta_diversity_through_plots.py –i ICF.biom –t ICF.tree –m ICF.mapping.txt –o results/beta_diversity –p params.txt –e 10267 –c Dx

This calculation will take from 1 - 5 min to complete.

Page 15: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 5: Taxonomy ComputationsFor this next step, we will create a graphical summary of the taxonomical composition of the samples.

Let us do the same thing as above, only this time merging the control and IC samples using the Dx column.

Results are located in ~/white/results/taxonomy (1st command) and ~/white/results/taxonomy_Dx (2nd command).

$ summarize_taxa_through_plots.py –i ICF.biom –m ICF.mapping.txt –o results/taxonomy

$ summarize_taxa_through_plots.py –i ICF.biom –m ICF.mapping.txt –o results/taxonomy_Dx –c Dx

Page 16: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 6: ANOVA TestsANOVA stands for Analysis of Variance. It is a standard suite of statistical tests aimed at explaining differences between groups of data.

We will use ANOVA in this step to see if there are any OTUs that explain the differences between sample categories.

We will use the ot_category_significance.py script in QIIME to do this.

The resulting file, ~/white/results/ANOVA.txt, sorts the OTUs in the data according to how likely they are driving the differences between samples.

The file includes probabilities (uncorrected and corrected), as well as abundance information and lineage of the OTU.

$ otu_category_significance.py –i ICF.biom –m ICF.mapping.txt –o results/ANOVA.txt –s ANOVA –c Dx

Page 17: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Statistical TestsIn this exercise, we will test our hypotheses. In particular, if the control and IC samples cluster together, the following tests will measure the significance of such clustering based on the metrics that we just calculated.

Page 18: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 7A: Statistical Tests - Diversity

In this step, we will see whether or not the IC and control samples differ significantly using the diversity results computed earlier.

We will use the compare_alpha_diversity.py script in QIIME to do this.

The result file is located in ~/white/results/species_significance.txt.

compare_alpha_diversity.py –i results/alpha_diversity/alpha_div_collated/observed_species.txt –c Dx –o results/species_significance.txt –d 10260 -m ICF.mapping.txt

Page 19: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 7B: Statistical Tests - Diversity

Let us take a look at the results file ~/white/results/species_significance.txt using the cat command.

It seems that the categories are very different.

We will confirm this later when looking at diversity plots

$ cat results/species_significance.txt

Comparison tval pvalControl,IC 3.65454556682 0.002

Output:

Page 20: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 8A: Statistical Tests - Diversity

In this step, we will see whether or not the IC and control samples differ significantly using the diversity results computed earlier. We will use the UniFrac matrix and the ANOSIM test.

We will use the compare_categories.py script in QIIME to do this.

The result file is located in ~/white/results/anosim/anosim_results.

$ compare_categories.py –-method anosim –i results/beta_diversity/unweighted_unifrac_dm.txt –m ICF.mapping.txt –c Dx –o results/anosim –n 9999

Page 21: Basic  Microbiome  Analysis with QIIME

Step 8B: Statistical Tests - Diversity

Let us take a look at the results file ~/white/results/anosim/anosim_results.txt using the cat command.

Although the p-value is significant, the R statistic says that the clustering is only moderately strong.

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

$ cat results/anosim/anosim_results.txt

Method name R statistic p-value Number of permutationsANOSIM 0.4694 0.0008 9999

Output:

Page 22: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

AnalysisWe will now analyze the files we generated during the and diversity runs and tests.

Page 23: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 9A: a Diversity Results

On the desktop, access our shared directory:

classes/mayo/white/results/

Inside the results directory, open the following file:

alpha_diversity/alpha_rarefaction_plots/rarefaction_plots.html

Select observed_species as metric, and Dx as category.

A graph will be displayed.

Page 24: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 9A: a Diversity Results

Control is significantly different than IC!

Page 25: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 10A: Diversity Results

On the desktop, access our shared directory:

classes/mayo/white/results/

Inside the results directory, open the HTML file in the following dir:

beta_diversity/unweighted_unifrac_2d_discrete

This will open a 2d PCA plot, based on unweighted UniFrac distances, colored by sample type (Dx, Control)

Page 26: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 10B: Diversity Results

Hover on the data points to obtain information about that sample…

Page 27: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Control and IC samples segregate, but only moderately. This is in agreement with the ANOSIM results (R=0.4694 , p = 0.0008).

Page 28: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Step 11: Taxonomy Results

On the desktop, access our shared directory:

classes/mayo/white/results/

Inside the results directory, open the HTML file in the following dir:

taxonomy/taxa_summary_plots/area_charts.html

Page 29: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

This is the taxonomy at phylum level, for all samples. Hover over each color to find out about each color (colors may differ from this plot).

These look like otherwise normal stool samples, with Firmicutes and Bacteroides dominating. Note the Fusobacteria in sample 2, a control!

Page 30: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Things get more complex as we go down the taxonomy hierarchy. This is the plot at genus level, typical of stool samples. There seems to be no obvious pattern, the usual case unless there’s something very wrong, or a known pathogen.

Hover over each color to see its taxonomy information.

Page 31: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Let’s see if there is something hidden in the taxonomy. In the results directory, open the ANOVA.txt file. OTU prob Bonferroni_corrected FDR_corrected Control_mean

IC_mean Consensus Lineage111 0.000113443547213 0.02507102393410.0250710239341 0.00310594468968 0.00022022007532

k__Bacteria; p__Bacteroidetes; c__Bacteroidia; o__Bacteroidales; f__Porphyromonadaceae; g__Odoribacter; s__unclassified22 0.00128127076471 0.283160839001 0.14158041950.0155471912415 0.00128661622402 k__Bacteria; p__Firmicutes; c__Clostridia; o__Clostridiales; f__Lachnospiraceae; g__unclassified; s__unclassified89 0.00148832607004 0.328920061478 0.1096400204930.00408471212469 0.000983251999578 k__Bacteria; p__Firmicutes; c__Clostridia; o__Clostridiales; f__Lachnospiraceae; g__Clostridium; s__unclassified154 0.0025315674133 0.559476398339 0.1398690995857.38470627331e-06 0.00183392914333 k__Bacteria; p__Tenericutes; c__Erysipelotrichi; o__Erysipelotrichales; f__Erysipelotrichaceae; g__Clostridium; s__Clostridium_ramosum

Odoribacter has 0.3% abundance in controls, 0.02% in IC…

Page 32: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Indeed, it seems to be a good marker despite its low relative abundance. Its absence seems correlated with IC (samples 4,7,8,9,10,12,14,15).

Page 33: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Analysis Conclusions

Microbial composition and structure significantly different in stool between IC patients and controls:

IC stool microbiota significantly less diverse

Overall IC microbiota different (it clusters away from controls)

Potential marker found:

Lack of Odoribacter associated with IC

Page 34: Basic  Microbiome  Analysis with QIIME

Basic Microbiome Analysis with QIIME v1 | Patrico Jeraldo and Bryan White

Exercise Conclusions

Basic microbiome analysis:

1. Calculate various diversity metrics for samples

2. Calculate statistical support for differences found between samples types

3. Plot taxonomy composition of samples

4. Basic tests for potential microbial markers