Elucidation of Signaling Pathways from Large-Scale Phosphoproteomic Data Using Protein Interaction Networks
Phosphoproteomic experiments typically identify sites within a protein that are differentially phosphorylated between two or more cell states. However, the interpretation of these data is hampered by the lack of methods that can translate site-specific information into global maps of active proteins and signaling networks, especially as the phosphoproteome is often undersampled. Here, we describe PHOTON, a method for interpreting phosphorylation data within their signaling context, as captured by protein-protein interaction networks, to identify active proteins and pathways and pinpoint functional phosphosites. We apply PHOTON to interpret existing and novel phosphoproteomic datasets related to epidermal growth factor and insulin responses. PHOTON substantially outperforms the widely used cutoff approach, providing highly reproducible predictions that are more in line with current biological knowledge. Altogether, PHOTON overcomes the fundamental challenge of delineating signaling pathways from large-scale phosphoproteomic data, thereby enabling translation of environmental cues to downstream cellular responses.
A zip archive containing several output tables can be downloaded from the results page of PHOTON.
subnet.csv: edges of the reconstructed signaling network. This is a sub-network of the input network (which is located at
subnet.csvfile can be e.g. imported into Cytoscape for further analysis.
scores.csv: signaling functionality scores devised by PHOTON. The 'Significant' column allows for easy filtering of the table. The other columns provide more detail on each step of the algorithm, i.e. calculating an empirical score, obtaining p-values, and finally FDR-corrected q-values.
go_scores.csv: details all GO annotation enrichment results for the reconstructed network. Filtering for the 'rejected' column (=1) will yield all sign. enriched categories. The naming of the other columns follows the convention of the statistical test used.
predictions.csv: contains information regarding the functional phosphorylation site prediction. The 'proba' column is the prediction score for each site. If the phosphorylation site was part of the training set the 'label' column will be 'TRUE'. One should take special care when the training set was small, i.e. few sites with 'label' set to TRUE are in the data set.
My result graph contains only a sigle node '-1'
Sometimes PHOTON will not yield any results. Known technical sources of failure are:
- The ANAT service is down. PHOTON relies on the ANAT web server which can be down for various reasons. Before reporting an error with PHOTON, please confirm that the Cytoscape plugin for ANAT is working as expected. Please note that you might still download the resulting activity scores derived by PHOTON.
Common errors when running PHOTON
Uploading a data file which is not
ASCIIencoded might yield errors such as
'utf-8' codec can't decode byte 0xa0 in position 0: invalid start byte
Converting the file to
ASCIIwill fix this issue.
There is a popup with a long error message similar to
<ns2:confidence>0.6912981534952782</ns2:confidence> <ns2:fromNodeId>473</ns2:fromNodeId> <ns2:toNodeId>126961</ns2:toNodeId> </ns2:edgesData> <ns2:edgesData> ...
This error messages is due to the ANAT web server not being available.
Please confirm that the Cytoscape plugin for ANAT is working as expected before reporting an error for PHOTON.
Please contact me or the authors of ANAT in case of errors.
Running PHOTON through Perseus
- Download Perseus from here.
- Install Python.
- Install perseuspy and PHOTON by running
pip install photon_ptm.
PluginPHOTONfrom the plugin store.
- Load the
Generic matrix upload.
- Filter the rows by
Confidence > 0.5.
- Create a network from the matrix.
- Load the experimental data.
- Annotate the nodes of the network with the experimental data.
- Run PHOTON from Network => Processing => Modifications => PHOTON.
- Use the full network for e.g. enrichment analysis. The reconstructed subnetworks can be
visualized. Signaling functionality scores are reported in a separate table and can be used
for clustering etc.
Running PHOTON on Docker
PHOTON runs inside a 'container' and therefore requires docker to run across all platforms.
General information regarding the installation of docker can be
found on the docker website.
Docker Toolbox (Windows, Mac)
First open the
Docker Quickstart Terminal. After initialization (can take some time),
denote the IP address of docker (under the whale image).
Now you can run PHOTON by entering:
docker run -d -p 5000:5000 jdrudolph/photon
Now you can access PHOTON from your browser under the IP address
of docker followed by colon and 5000. For example
Alternatively you can lookup the IP address using
docker-machine ip default.
Docker for Windows / Linux
Open a Powershell (terminal in linux) and run:
docker run -d -p 5000:5000 jdrudolph/photon
Now you can access PHOTON from your browser.
Native Linux (Experts)
Dockerfile on how to run PHOTON on Linux natively.
First list all running containers:
Lookup the name of the container (last column) and stop it using
docker stop name_of_container
New versions of docker can be downloaded by running
docker pull jdrudolph/photon
before issuing the
run command as shown above.
Please open an issue here, or
contact me directly if you have issues with installing or using PHOTON.
If the current licencing (see
LICENCE.txt) does not suit your needs,
please contacte me for alternative licencing options.
PHOTON data format example
GeneID,Amino.Acid,Position,avg,Symbol 5097,S,962,-0.47417884405658556,PCDH1 5097,S,984,-0.6438018715183813,PCDH1 4300,S,522,-0.5172999908818245,MLLT3 81628,S,264,0.8682687511988673,TSC22D4 81628,S,279,2.684096205546553,TSC22D4 4034,S,25,-0.593695490783283,LRCH4 4034,S,380,1.6585730683114723,LRCH4 6651,S,1822,0.046145203537912814,SON 6651,S,1737,-0.06153405253207198,SON