Genboree 16S Workbench Workshop Part I

Report
Genboree Microbiome Workbench
16S Workshop Part I
March 11th, 2014
Julia Cope
Emily Hollister
Kevin Riehle
•
•
•
•
•
•
•
•
•
Genboree Workflow
Create Group
Create Database
Create Project
Upload Files 
Create Samples (Sample Import using metadata
file) 
Link Samples to Sequence Files (Sample File
Linker) 
QC and Attach Sequences (Sequence Import) 
QIIME   
RDP 
Data Analysis - QIIME
• How to select samples for analysis
• Chimera removal and why you should be
thinking about it
• Output
– downloading and organization
– making sense of the files
Data Analysis - QIIME
• How to select samples for analysis
Data Analysis - QIIME
– Selecting samples for analysis
• INPUT = One or more Sequence Import folders
– All should be of the same variable region; ideally produced with the same primer and
sequencing direction
• OUTPUT Targets = Your database (required), your project (optional)
Data Analysis - QIIME
Caveats:
• All samples in your input folder will be analyzed
– This includes no-template controls and positive controls
– The % variation explained by you PCoA may be influenced by the
inclusion of these samples
• QIIME on Genboree is not currently set up to allow users to subsample
their data
– This can be problematic if sequencing depth varies substantially
across samples
– It does however perform a “rounding up” normalization step
A bit about sequencing depth
How deep should you go?
There is no good answer
Strong biological patterns can be detected
with low sequencing depth
– 10s to 100s of sequences can
sometimes be enough
– 1000s tend to be the norm
Subtle biological patterns tend to require
greater sequencing depth for detection
Sequencing depth can be dictated by:
– Sample quality
– The number of samples placed on a run
– Project budget
Kuczynzski et al. 2010 Nature Methods 7: 813-819
Unequal sequencing depth
What’s the problem?
Being certain that you are seeing the full view
(…or at least equivalent glimpses of the) of your communities
http://www.cs.unc.edu/~lguan/Research.files/backgroundSubtractionResult.JPG
Unequal sequencing depth
What’s the problem?
Unequal depth
Avg Red = 5995 seqs
Avg Blue = 11672 seqs
Same data set
Sampled are colored
by library size
Red ~4000
Orange ~5000
Yellow ~6000
Green 8,000-10,000
Blues 11,000-17,000
Unequal sequencing depth
What’s the problem?
Unequal depth
Avg Red = 5995 seqs
Avg Blue = 11672 seqs
Equal depth
All libraries were
sub-sampled to
~4000 reads.
Data Analysis - QIIME
• Chimera removal and why you should be
thinking about it
– What is a chimeric sequence?
– How frequently do they occur?
– An example from real data
– Why should you think about chimeras?
– How to screen for chimeras using Genboree
What is a Chimeric Sequence?
– In Greek mythology:
• A creature that was an amalgam of
multiple animals
• Body of a lion, head of a goat, tail
resembling a snake
– In your sequence data:
• The combination of multiple sequences
during PCR to create a hybrid
– In sequence databases:
• A not-so-small nightmare of junk data
• Mis-annotation
• Enhanced “discovery” of novel organisms
Chimera generation figure from: Haas et al. 2011, Genome Research 21:494-504
How frequently do chimeras occur?
– Schloss et al 2011:
• With mock communities of known
composition:
• ~8% of raw sequences were chimeric
• Incidence increased with sequencing depth
Likely Chimera
Parent 1
AATCGCGACCTGTTTAACCGTAGGTC
Query
AATCGCGACCTGTGCTACACGGGTA
Parent 2
AAACGCTTACGGAGCTACACGGGTA
Non-chimera
Parent 1
AATCGCGACCTGTTTAACCGTAGGTC
Query
AATCGCGACCTGTTTAACCGTAGGTC
Parent 2
AAACGCTTACGGAGCTACACGAGTC
Schloss et al. 2011 PLoS ONE 6(12):e27310
– Approaches for detection:
• Multiple algorithms available
• Genboree uses ChimeraSlayer
– How it works:
• The ends of each read (~30% of total length)
are compared to a chimera-free reference
database
• Potential “parent” sequences are identified
• Identity of potential chimera to in silico
chimera evaluated
An example from real data
Alignment of chimeric sequences derived from Streptococcus (top, red) and Staphylococcus (bottom, black)
Sequences were generated from 4 replicate PCR reactions/454 runs of V3V5 sequence
Chimeric alignment from: Haas et al. 2011, Genome Research 21:494-504
Why should you think about chimeras?
– Spurious results
• Artificially increases estimates of richness and
diversity
• You may discover a “new” (but fake) species
– Should you trust all flagged chimeras?
• Most people do but….buyer beware
• False-positive rates are in the 1-4% range
• Some taxa are poorly represented in reference
databases
• Prevotella and Acinetobacter are known to produce
false-positive results in ChimeraSlayer
– How to verify (digging in to your QIIME output)
• Obtain representative sequence(s) and verify their
identity (e.g., BLAST vs. NCBI nt database, RDP
SeqMatch)
Sogin et al 2006 PNAS 103:12115-12120
How to screen chimeras in Genboree
– Run a QIIME job
• INPUT = Sequence Import folder
• OUTPUT Targets = Your database (required), your project (optional)
How to screen chimeras in Genboree
– Select “Remove Chimeras” in the Tool Settings dialogue box
• Provide a study name
• Provide a job name (TIP: add chimeras_removed to you job name so that
your output reflects that you selected this option)
• Click SUBMIT
Data Analysis - QIIME
• Output
– downloading and organization
– making sense of the files
How do I get my files out?
– Entire folders can be archived/downloaded
• INPUT = Folder to be archived
• OUTPUT = Database to house archive
How do I get my files out?
– Entire folders can be archived/downloaded
• Provide and archive name
• Choose your compression type
• Decide if you want the directory structure to be preserved
• SUBMIT
How do I get my files out?
– Single files, including archives, can be downloaded one by one
• Click on your file of interest in the DATA SELECTOR window
• Click on the “Click to Download File” link in the DETAILS window
• Save the file to your computer or storage drive
• Most file types will require decompression
QIIME – making sense of the files
–
–
–
–
–
–
–
–
–
–
–
fasta.result.tar.gz
jobFile.json
mapping.txt
otu.table
phylogenetic.result.tar.gz
plots.result.tar.gz
raw.results.tar.gz
repr_set.fasta.ignore
sample.metadata
settings.json
taxonomy.result.tar.gz
QIIME – making sense of the files
– fasta.result.tar.gz: multiple sequence alignment of your representative sequences file.
Rep seqs = representative sequence for each OTU.
– jobFile.json: a log of the settings used by Genboree to run your analysis
– mapping.txt: a QIIME-compatible metadata file, includes barcode information
– otu.table: a spreadsheet of OTU by sample distributions
– phylogenetic.result.tar.gz: a phylogenetic tree of your rep seqs, additional files
required for iTOL
– plots.result.tar.gz: figures, html files for all PCoA plots produced in your QIIME run
– raw.results.tar.gz: mapping file, otu table, rep seqs file, distance matrices underlying
all PCoA calculations
– repr_set.fasta.ignore: RDP classification (with confidence scores) of each rep seq
– sample.metadata: like the mapping.txt file, with additional file locations for Genboree
– settings.json: similar to the jobFile.json file
– taxonomy.result.tar.gz: taxonomic summaries (per sample, at the Kingdom, Phylum,
Class, Order, Family, and Genus levels)
•
•
•
•
•
•
•
•
•
Genboree Workflow
Create Group
Create Database
Create Project
Upload Files 
Create Samples (Sample Import using metadata
file) 
Link Samples to Sequence Files (Sample File
Linker) 
QC and Attach Sequences (Sequence Import) 
QIIME   
RDP 
Data Analysis - RDP
• How to select samples
• Output
– Downloading and organization
– making sense of the files
Data Analysis - RDP
– Selecting samples for analysis
• INPUT = One or more Sequence Import folders
– All should be of the same variable region; ideally produced with the same primer and
sequencing direction
• OUTPUT Targets = Your database (required), your project (optional)
Data Analysis - RDP
Caveats:
• All samples in your input folder will be analyzed
– This includes no-template controls and positive controls
• RDP on Genboree does not pre-filter for chimeric sequences
• RDP on Genboree is not currently set up to allow users to subsample their
data
– Depending on your application, this may be problematic if sequencing
depth varies substantially across samples
– It does however perform a “rounding up” normalization step and
presents data on a relative abundance basis
How do I get my files out?
– Entire folders can be archived/downloaded
• INPUT = Folder to be archived
• OUTPUT = Database to house archive
How do I get my files out?
– Entire folders can be archived/downloaded
• Provide and archive name
• Choose your compression type
• Decide if you want the directory structure to be preserved
• SUBMIT
How do I get my files out?
– Single files, including archives, can be downloaded one by one
• Click on your file of interest in the DATA SELECTOR window
• Click on the “Click to Download File” link in the DETAILS window
• Save the file to your computer or storage drive
• Most file types will require decompression
RDP – making sense of the files
–
–
–
–
–
–
–
–
–
–
–
–
–
–
domain.result.tar.gz
phylum.result.tar.gz
class.result.tar.gz
order.result.tar.gz
family.result.tar.gz
genus.result.tar.gz
sample.metadata
settings.json
count.result.tar.gz
count.xlsx
count_normalized.xlsx
weighted.xlsx
weighted_normalized.xlsx
png.result.tar.gz
RDP – making sense of the files
–
–
–
–
–
–
–
–
–
–
–
–
–
domain.result.tar.gz
phylum.result.tar.gz
class.result.tar.gz
Per sample summaries at various taxonomic levels, including raw
order.result.tar.gz
counts and weighted values
family.result.tar.gz
genus.result.tar.gz
sample.metadata
settings.json
count.xlsx
Per sample summaries at various taxonomic levels, raw counts or
relative abundances (normalized)
count_normalized.xlsx
weighted.xlsx
Per sample summaries at various taxonomic levels, weighted by
confidence of ID assignments (raw counts or normalized)
weighted_normalized.xlsx
All of the plots produced during your run (e.g., heatmaps, stacked bar graphs)
png.result.tar.gz
Individual Time
• Confirm user accounts are created.
• Confirm users know where mock data or their
data set are.

similar documents