metaSpectraST is an unsupervised and database-independent analysis tools for metaproteomic MS/MS data using spectrum clustering. It clusters all experimentally observed MS/MS spectra based on their spectral similarity and create a representative consensus spectrum for each cluster by using the spectrum clustering algorithm implemented in the spectral library search engine, SpectraST.
Spectrally similar MS/MS spectra that are grouped in one spectral cluster are presumed to originate from the same peptide sequence, and therefore metaSpecraST treats them as replicate spectra and quantitatively profiles samples by counting the number (spectral count, SC) or intensity (spectral index, SIN) of replicate spectra in each spectral cluster.
The metaSpectraST spectral clusters also offer a portal to integrate and reconcile multiple peptide identification approaches, including database search, open modification search, and de novo sequencing. For each spectral cluster, sequences of raw spectra and their cosnensus spectrum assigned by different identification methods vote for the consensus peptide sequence of the spectral cluster through a heuristic reconciliation scheme and the majority rule.
With metaSpectraST you can,
- Fast profile and compare the microbial communities of your sample;
- Classify your metaproteomic (or proteomic) samples;
- Validate biological/technical replicates;
- Integrate and reconcile multiple peptide/protein identification approaches for further taxonomic or functional studies.
- Python version >= 3.7, R version 4.1.3
- SpectraST (v5.0)
SpectraST is an integral component of the Trans Proteomic Pipeline suite (TPP) of software. A compiled executable file is included here, which can be used alone without other TPP components.
We encourage users to download and install the entire TPP suite, which provides other useful functionalities such as raw data importation, automatic validation of search results, protein inference, and quantification and visualization. Please refer to the guides for TPP Linux installation, and the official download site for Windows installer.
- edgeR (v3.34.0)
metaSpectraST normalizes the data using the trimmed mean of M-values (TMM) normalization method implemented in the edgeR package. edgeR is not necessary if you would like to normalize the data with other methods. Please refer to Bioconductor-edgR for further information.
To install the edgeR package, start R and enter:
if (!require("BiocManager", quietly = TRUE))
install.packages("BiocManager")
BiocManager::install("edgeR")You may also need to install the limma (v3.48.1) package
if (!require("BiocManager", quietly = TRUE))
install.packages("BiocManager")
BiocManager::install("limma")- Download or clone the repository:
git clone https://github.com/bravokid47/metaSpectraST.git- Make metaSpectraST executable by adding the directory
yourpath/metaspectrast/to the environment variable$PATH, or just copy the following line to the~/.bashrcor~/.bash_profilefile andsourcethe file.
export PATH="$PATH:yourpath/metaspectrast";metaSpectraST can perform spectral clustering from the following data formats:
- mzML format
- mzXML format
- mgf format
Please note that mgf format is required for computing the normalized spectral index (SIN). File formats can be converted with msconvert or ThermoRawFileParser.
There are 6 individual modules in metaSpectraST. Run the following command to get explanation of the 6 modules.
metaspectrast -hOutput
>>>
_________ metaSpectraST by Hao, Chunlin _________
metaSpectraST v=0.0
Usage: metaspectrast [module]
Module:
1 cluster Clustering MS/MS spectra and create consensus spectra
2 computesc Spectral count-based (SC) sample profiling
3 computesin Normalized spectral index (SIn) based sample profiling
4 normalize Normlizing the data matrix of sample profiles (SC or SIn)
5 classify Hierarchically clustering and classifying samples
6 reconcile Reconciliation scheme
Each module is run separately. For example, to run the computesin module,
metaspectrast computesin -hOutput
>>>
usage: metaSpectraST_SIn.py [-h] [-s [SPTXT]] -m MGF [MGF ...]
metaSpectraST (v0.0) by Hao, Chunlin.
Compute normalized spectral index (SIn) of consensus spectra.
optional arguments:
-h, --help show this help message and exit
-s [SPTXT] consensus spectra .sptxt file, grandConsensus.sptxt by default.
-m MGF [MGF ...] raw spectra data sets in MGF format
Run the following command to perform spectral clustering:
metaspectrast cluster <path/*mzML>Fragmentation type (ETD, HCD, CID-QTOF) of the spectra can be specified by the -i option. Default is off and the fragmentation type can be determined from the data files.
metaspectrast cluster -i HCD <path/*mzML>When this step is done, it produces three types of output file in the working directory. The file bar.splib is the spectra library in a binary format. The bar.sptxt is a human-readable version of the bar.splib. The files bar.spidx and bar.pepidx are indices on the precursor m/z value and peptide, respectively. The file grandConsensus.sptxt is the library of consensus spectra, which will be used in the subsequent steps. A library of consensus spectra in .mgf format is also produced, named as grandConsensus.mgf.
Consensus spectrum created in step 1 can be quantified by counting the number (spectral count, SC) or intensity (spectral index, SIN) of the replicate spectra (raw spectra) in the corresponding spectral cluster in the sample. Quantified consensus spectra can then be used to profile the samples.
Spectral count-based (SC) profiling
metaspectrast computesc -s <path/grandConsensus.sptxt>When it is done, it produces two CSV files, unnorm_consensusPep_SC.csv and consensusSpec_RawSpectra_idx.csv. The file unnorm_consensusPep_SC.csv is unnormalized spectral count of consensus spectra in each sample, which can be normalized by the normalize module (see Step 3) or simply normalized by the sum of the spectral count in each data set. The file consensusSpec_RawSpectra_idx.csv is the index of the correspondence of raw spectrum and its consensus spectrum.
Normalized spectral index-based (SIN) profiling
metaspectrast computesin -s <path/grandConsensus.sptxt> -m <path/*mgf>Note that the .mgf file has to be named the same as the the corresponding input file in Step 1.
When it is done, it produces three CSV files, unnorm_consensusPep_SI.csv, consensusPep_SIn.csv and consensusSpec_RawSpectra_idx.csv. Similar to SC profiling, the file unnorm_consensusPep_SI.csv is unnormalized spectral index of consensus spectra in each sample, which can be normalized by the normalize module (see Step 3). The file consensusPep_SIn.csv is the same file as unnorm_consensusPep_SI.csv, but normalized by the sum of the spectral index in each data set. The file consensusSpec_RawSpectra_idx.csv is the index of the correspondence of raw spectrum and its consensus spectrum.
Hierarchical clustering of samples can be performed based on their SIN or SC profiles. But before that, SIN or SC profiles need to be normalized.
Normalization of the SIN or SC profiles can be performed by simply diving by the sum of the SIN or SC in each data set. Note that the file consensusPep_SIn.csv can be used for hierarchical clustering directly as it has been normalized by the sum of the SIN in each data set already. Alternatively, run the following command to normalize the SIN or SC profiles. The normalize module normalizes the data using the trimmed mean of M-values (TMM) normalization method implemented in the edgeR package.
metaspectrast normalize -u unnorm_consensusPep_SI.csvThe normalized file is named as tmmNorm_consensusPep.csv. The scaling factor of each data set can be found in the file tmm_scalingFactor.csv.
Run the following command to perform the hierarchical clustering on samples.
metaspectrast classify -n tmmNorm_consensusPep.csvIf you are working with microbial communities in or on a host organism, such as gut microbiome, you could exclude the consensus spectra of the host peptides by providing metaSpectraST a .txt file containing a single column of the full name of the consensus spectra that need to be excluded. Consensus spectra of host peptides can be identified by database search of the consensus spectra library (grandConsensus.mgf) against the host proteome database.
metaspectrast classify -n tmmNorm_consensusPep.csv -r removal.txtFurther refinement of the SIN or SC profiles can be conducted through the -o option, which specifies the minimum number of samples that a consensus spectrum has to be observed in. This option can help filter out the singly or rarely observed consensus spectra. Default is 1.
metaspectrast classify -n tmmNorm_consensusPep.csv -r removal.txt -o 3At the end of the day, two figures are produced. One is the file hierarchicalHeatmap.png, which is a heatmap with some decorations showing the clusters of the samples. The other is the file dendrogram.png, which is a dendrogram of the clusters of the samples.
Peptide sequences of raw spectra and consensus spectra can be identified through multiple peptide identification methods, including database search, open modification search, and de novo sequencing. Replicate spectra and the consensus spectrum in one spectral cluster may be identified as conflicting peptide sequences by different identification methods. The reconcile module employs a heuristic reconciliation scheme to resolve the conflicts and correct sequence errors.
Reconciliation scheme
Preparing identification results
To run the reconcile module, you need to prepare the identification results first. It's very simple. You just need to organize the identification results of raw spectra as CSV files containing at least two columns, RawSpectrum and peptide. The 'RawSpectrum' column lists the recorded name of each raw spectrum, while the 'peptide' column lists the peptide sequence assigned to each raw spectrum. If the identification method provides protein information, you can add it as an additional column with the header of protein. Identification results from different identification methods should be organized as separate CSV files. For identification results of consensus spectra, please change the header RawSpectrum to consensusSpec.
Then run the following command to start the reconciliation procedure.
metaspectrast reconcile -d raw_DB.csv -t raw_taggraph.csv -n raw_denovo -c consensus_DB.csv -i consensusSpec_RawSpectra_idx.csvOption arguments
-d Specifies the result of database search of raw spectra.
-t Specifies the result of open modification of raw spectra.
-n Specifies the result of de novo sequencing of raw spectra.
-c Specifies the result of database search of consensus spectra.
-i Specifies the index of correspondence of raw spectrum and its consensus spectrum.
By default, 'consensusSpec_RawSpectra_idx.csv' generated automatically by the 'computesc' or 'computesin' module will be used.
Author: Hao, Chunlin
Principle Investigators: Henry H. N. Lam (HKUST), Patrick K. H. Lee (CityU), and Joshua Elias (CZ Biohub).
Hao, C., Elias, J.E., Lee, P.K.H. et al. metaSpectraST: an unsupervised and database-independent analysis workflow for metaproteomic MS/MS data using spectrum clustering. Microbiome 11, 176 (2023). https://doi.org/10.1186/s40168-023-01602-1 https://microbiomejournal.biomedcentral.com/articles/10.1186/s40168-023-01602-1#citeas