Here are the steps I followed for the Montipora capitata coral project after running the samples on the Mass Spectrometer (MS).
Each heading represents a distinct step in the analysis and under each heading there is a description of what, why and how the analysis was done, and includes where to find the files used in each analysis and where the output is. For much of these earlier analyses, the Putty command line was used to access the UW genome sciences department where the raw data is stored and where programs were run in their department's cluster. As of yet, I do not know how to download these files to my computer, nor do I have space on my computer to store the larger files.
Directories are italicized
Commands are in bold
Prior to running the MS, there were a few issues with sample preparation.
- Sample 69B, both inner and outer, had very low protein concentration.
- Half of the samples were desalted incorrectly. I misread the desalting protocol added 100ul of sample instead of 100ug (all of the sample), which accounted for about 3/4 of the entire sample so the next day, the remaining 1/4 of the samples were desalted and added to make the samples whole. To QC these samples, they were evaluated comparing their protein counts to correctly desalted samples and visually compared in NMDS plots from Abacus outputs.
- During MS analysis, the LC column broke-down and protein counts were low for about 1/3 of the samples. The column was replaced and those samples were ran again. Those samples were QC'd the same way as explained above.
A comet search was run on all MS generated .raw files. The comet search was run using Putty command line via the UWPR cluster in the UW department of Genome Sciences. This search compares the proteins found in the samples to the genome for Montipora capitata to quantify which proteins were being expressed at the time of sampling. The comet searches were run three times: once for inner matrix samples, once for outer matrix samples and once for Rerun samples (the Reruns were reran because the column started to break down halfway through MS analysis, so we replaced it and reran samples that were not looking well).
The comet parameters can be found here:
nexus1 /net/nunn/vol1/brook/2018_March_9_JPG_corals/comet.params
They are the same parameters used for all three searches.
The Montipora genome data base used for the searches can be found here (directory is also noted in the comet.params file):
nexus1 /net/gs/vol4/shared/nunnlab/search/emmats/databases/Montiporacapitata.contam.fasta
To convert the .raw files to .mzxml and .pep.xml files and run the comet search, the following directories were opened and these commands were used:
outer matrix: nexus1 /net/nunn/vol1/jeremy/2018_March_9_JPG_corals/coral_outer
convert.sh *.RAW
runCometQ *.RAW
Inner matrix: nexus1 /net/nunn/vol1/brook/2018_March_9_JPG_corals/coral_inner_matrix
convert.sh *.RAW
runCometQ *.RAW
Rerun inner matrix: nexus1 /net/nunn/vol1/brook/2018_March_9_JPG_corals/coral_inner_matrix/Rerun_SOM_samples
convert.sh *.RAW
runCometQ *.RAW
Peptide- and Protein-prophet were used to validate the quality of the peptides and proteins identified in Comet searches. Additionally, peptide and protein quantities are given during the anlysis allowing us to QC some of the issues we had with sample prep (see above). To run the prophets, xinteract files were generated using Putty and Comet programs, see below to follow the steps.
To run the prophets, the following directories were opened:
outer matrix: .mzXML and .pep.xml files cannot be found at this time
inner matrix: .mzXML and .pep.xml files are found here: nexus1 /net/nunn/vol1/brook/2018_March_9_JPG_corals/coral_inner_matrix
Rerun inner matrix: .mzXML and .pep.xml files cannot be found at this time
Then this commands was used (while in each directory) to generate interact-.pep-MODELS.html, interact-.pep.xml, interact-.pep.xml.index, interact-.prot.MODELS.html, interact-.prot.xml" files:
runCometQ --wocomet --single *.mzXML
And this command was used to generate an interact-COMBINED.prot.xml file:
runCometQ --wocomet --all *.mzXML
These outputs (stored in the same directories as the input files-see above) opened a separate protein- or peptide-prophet web page for each sample with the suffix: .xml. For each sample, the results give the the probability of a good match with proteins and peptides predicted to occur from the M. capitata genome, and give homologous scores for comparisons with decoys. They also give the numbers of unique proteins and peptides.
A list of peptide and protein counts for the samples can be found here:
~/Documents/GitHub/Jeremy-coral/data/protein\ counts.xlsx, or here
Abacus was run on all coral samples to extract adjusted spectral counts to be used to visualize and quantify protein data using R between different experimental groups (e.g. inner, outer, bleached and non-bleached).
Because Abacus attempts to grab all files in a directory, all interact-.pep.xml, interact-.prot.xml and the interact-COMBINED.prot.xml were comliled into this single directory:
nexus1 /net/nunn/vol1/brook/2018_March_9_JPG_corals/All_coral_samples
The Abacus parameters were also adjusted and added to the same directory, here:
nexus1 /net/nunn/vol1/brook/2018_March_9_JPG_corals/All_coral_samples/Abacus_parameters.txt
Then, the following command was used to run Abacus:
java -Xmx16g -jar /net/pr/vol1/ProteomicsResource/bin/abacus.jar -p Abacus_parameters.txt
The output was downloaded from the UW genome cluster (not sure how- need to ask Emma) and can be found here:
~/Documents/GitHub/Jeremy-coral/data/ABACUS_output.tsv, or here
A .csv was also generated (not sure how again) for use in R, and can be found here:
~/Documents/GitHub/Jeremy-coral/data/Abacus_output.csv or here
To visually compare differences in protein expression between coral treatments (bleach, non-bleached and inner, outer tissue) NMDS plots were generated in R. The code used was provided by Emma Timmins-Schifman. It essentially does a cluster analysis based on similarity of protein content in each sample based on these metrics:
- number of unique peptides: NUMPEPSUNIQ
- spectral count: NUMSPECSTOT
- adjusted spectral count: ADJNSAF
An initial set of analyses were run to QC some prep issues (see above). The r scripts, .csv files, and figures can all be found here:
~/Documents/GitHub/Jeremy-coral/scripts/R/R_scripts_Abacus_analysis link ~/Documents/GitHub/Jeremy-coral/scripts/R/csv_files_Abacus_analysis link ~/Documents/GitHub/Jeremy-coral/scripts/R/figures_Abacus_analysis link
From these analysis, I was able to distill the data down to the best data, deciding to remove the duplicates with lower protein counts, the samples where the column degraded and keep the rerun samples. Also the incorrectly desalted samples seem OK. Sample 69B looks like an outlier but is kept with the newly named all good samples for now.
From these all good samples data, I am starting to look for general differences between bleached and non-bleached, and inner and outer samples. The R project for these analysis can be found here:
~/Documents/GitHub/Jeremy-coral/scripts/R/R_project_Abacus_analysis link
In that directory is the R project, titled R_project_Abacus_analysis including the .csv files for all bleached, all non-bleached, all inner, all outer, and the original Abacus output.
The figures were output here:
~/Documents/GitHub/Jeremy-coral/scripts/R/figures_Abacus_analysis link
And here there are:
It looks like the most interesting comparison will be between inner and outer samples. As of yet, there does not appear to be any distinguishable patter between bleached and non-bleached samples.
I need to talk to Brook about the inner sample figure where a warning message was generated when using the metaMDS function:
Warning message: In metaMDS(Mcap.tra, distance = "bray", k = 2, trymax = 100, autotransform = F) : stress is (nearly) zero: you may have insufficient data
Eigenvectors help show which proteins are driving the differences between cluster groups in the nmds plots. They were calculated in R using the envfit function in the vegan package. To see the code used to produce this, scroll towards the bottom of any of the R scripts for inner, outer, bleached or non-bleached here.
Eigenvectors were displayed visually on nmds plots here. It is difficult to tell which proteins are represented on the plots, so lists of significant eigenvectors were made as .txt files at < 0.01 p-values and < 0.001 p-values, found here.
As there was a clear x-axis division between inner and out samples, the eigenvector lists will be used to identify the proteins pulling the cluster grouping towards the inner samples (right, positive values) and outer samples (left, negative values).
Qspec is another way to determine significant protein differences between treatment groups. It takes the number of unique proteins found in each sample from two treatments, a control and a treatment, and compares them giving a list of significantly up-regulated proteins in the treatment group. For these analyses, non-bleached or outer tissue were the controls and bleached or inner tissues were the treatments.
The format for qspec input files was followed using these instructions, based on files generated from the Abacus output and the same R scripts found here.
Those files were saved in this repository here and this directory in the UWPR GS cluster:
nexus1 /net/nunn/vol1/jeremy/2018_March_9_JPG_corals/qspec
Qspec was ran in Putty through the tephra node of the UWPR genome sciences cluster by navigating to the directory shown above and using these command line prompts:
ssh tephra
qspec-param Mcap_inner_spec_counts_for_qspec.txt 2000 10000 0> cat Mcap_inner_spec_counts_for_qspec.txt_qspec
qspec-param Mcap_outer_spec_counts_for_qspec.txt 2000 10000 0
qspec-param Mcap_bleached_spec_counts_for_qspec.txt 2000 10000 0
qspec-param Mcap_non_bleached_spec_counts_for_qspec.txt 2000 10000 0
The output files end with the suffix .txt_qspec and are found here
Uniprot is an online resource of protein sequence and annotation data. We used uniprot to compile a suite of information about the proteins observed in our samples, including:
- protein names
- gene names
- Organism (that the protein is associated within this case mostly Stylophora pistillata)
- Gene ontology (biological process)
- Gene ontology (cellular component)
- Gene ontology (GO)
- Gene ontology (molecular function)
- Gene ontology IDs
- Cross-reference (KEGG)
- Cross-reference (KO)
- Cross-reference (Pfam)
To gather this imformation, the entire list of proteins from the fasta results we uploaded on the Retrieve/ID mapping page. The from "UniprotKB AC/ID" to "UniprotKB" option was selected, then the list was submitted. After that, The info listed above was selected and downloaded which allowed us to identify 'most' proteins in our samples and continue with Gene Ontology analysis as well as begin to figure out what is going on functionally in each treatment.
MetaGomics is a web-based tool that helps visualize protein functions among treatment groups compared to the the known proteome (in this case the "proteome" for M. capitata determined by MS). The fasta results for M. capitata were sent to Mike Riffle via this website searching the "uniprot sprot" database with a cutoff of 1E10. M. Riffle then set up this webpage where protein lists can be entered and a map and raw data of Gene Ontology function are generated.
The generated text files can be found here
The maps are currently of very low quality and not in a format I am familiar with. They can easily be generated again by entering the significant protein list for each treatment here.
DAVID is an online functional annotation tool that work similar to MetaGOmics but requires annotated "omics" information to work. We thought that since we could find annotations via the uniprot platform that we would be able to use DAVID, but for now we are not
Venn diagram showing number of proteins unique to each treatment and to every combination of treatments (bleached, non-bleached, inner and outer tissue)
The diagram was made by submitting lists of proteins from each treatment to Venny an online platform for creating Venn diagrams
Below is a final NMDS ordination plot showing how all treatment types clustered using Bray dissimilarity index. The vectors represent a subset of significant (p < 0.001) proteins that influenced the ordination pattern. Code for making this figure can be found here
