SLIDE 1
Cbio 16S analysis pipeline
Katie Lennard
Microbiome analysis workflow
Data preprocessing (UCT High Performance Cluster)
Cbio 16S analysis pipeline Katie Lennard Microbiome analysis - - PowerPoint PPT Presentation
Cbio 16S analysis pipeline Katie Lennard Microbiome analysis - - PowerPoint PPT Presentation
Cbio 16S analysis pipeline Katie Lennard Microbiome analysis workflow Data preprocessing (UCT High Performance Cluster) Microbiome analysis workflow unsupervised classification correlations analyses Import data into R Microbiome analysis
SLIDE 2
SLIDE 3
Microbiome analysis workflow
correlations analyses unsupervised classification
Import data into R
SLIDE 4
Microbiome analysis workflow
correlations analyses unsupervised classification
Exploratory
Summary barplots
SLIDE 5
Microbiome analysis workflow
correlations analyses unsupervised classification
Exploratory
Beta diversity: NMDS/PCoA
SLIDE 6
Microbiome analysis workflow
correlations analyses unsupervised classification
Exploratory
Annotated heatmaps
SLIDE 7
Microbiome analysis workflow
correlations analyses unsupervised classification Differential abundance testing
Downstream analyses
SLIDE 8
Microbiome analysis workflow
correlations analyses
Downstream analyses
SLIDE 9
Microbiome analysis workflow
correlations analyses unsupervised classification unsupervised classification
Downstream analyses
SLIDE 10
Microbiome analysis workflow
correlations analyses unsupervised classification Biomarker discovery: random forests
Downstream analyses
SLIDE 11
Customized .R script to make your life easier
- Convert from phyloseq object to metagenomeSeq object
- Get the lowest available taxonomic annotation for each OTU and merge
counts at this level
- Heatmap (using NMF package) customized for phyloseq objects
- Can easily specify a subset of taxa and/or samples to plot
- Select annotation colours
- Select distance function for clustering
- Choose to merge taxa at a given level (e.g. Genus) or plot individual OTUs
- Generic barplot function build on phyloseq plot_bar()
- Specify subset of samples
- Filter OTUs so very rare ones (that just clog up the legend) are excluded
- Merge at any taxonomic level (Family, Genus etc..)
- Differential abundance testing + heatmap of significant results
- Built around MetagenomSeq’s fitzig() and mrfulltable() functions
- NB: currently only setup for two-class categorical comparisons
- Correlations testing + correlation plot of significant results
SLIDE 12
Customized .R script to make your life easier
- For PICRUSt data: takes the output from PICRUSt's
metagenome_contributions.py, together with taxonomic annotation for the OTUs included in this table and provides a summary of the contribution of each Family/Genus.. etc to ONE SPECIFIC KEGG gene e.g. K02030
- Random forests analysis on the otu table of a supplied phyloseq
- bject
- The data is randomly divided into a training (two thirds of the data) and test
set (remaining one third of the data not used for training)
- Results printed to screen and written to file including:
- most important taxa, AUC, PPV, NPV, OOB errors, class errors
- option to specify the top N taxa to see how they perform
SLIDE 13
Random Forests output example
SLIDE 14
Random Forests output example
SLIDE 15
The 16S accreditation dataset: first look
- Number of OTUs: 181 (140 retained after filtering)
- Number of samples: 15
- Sample data summary (columns=Treatment; rows=Dog):
0 1 2 3 4 B 1 1 1 1 1 G 1 1 1 1 1 K 1 1 1 1 1
SLIDE 16
The 16S accreditation dataset: first look
SLIDE 17
The 16S accreditation dataset: first look
SLIDE 18