PPIXpress

Protein interaction data Generally, transcripts or genes in all files (with the exception of the TCGA formats) are assumed to be given as Ensembl identifiers. Proteins are assumed to be UniProt accessions or HGNC/Ensembl genes in the input network. Furthermore, compressed files should have the ending .gz.

• From file
  PPIXpress accepts protein interaction data of Simple Interaction Format (SIF). SIF is a convenient format to describe a graph as a list of interactions.
  PPIXpress expects interacting proteins pairs (UniProt accessions, Ensembl genes or HGNC gene symbols) line-by-line and optionally a weight.
  A header is not necessary, if there is one it should be of the following form:

  Protein 1	Protein 2	weight
  Q9NUQ3	Q9UNE7	1.0
  Q9NXF1	Q9UL18	1.0
  ...	...	...

  Example: A protein-protein interaction network of physical interactions in human merged from data of IntAct (release 186), BioGRID (3.2.120) and HPRD (r9) example_ppi_data.sif.
• From web
  Alternatively a network can be automatically retrieved from the current Mentha or IntAct databases.
  Example: Type 9606 for a network in human.

Expression data

• Cufflinks (isoform/gene) FPKM tracking files Cufflinks (V2.x) creates FPKM tracking files with annotations according to a specified genome-annotation GTF file.
  If Ensembl/GENCODE annotation is used, those files can be immediately used by PPIXpress.
  Example: A FPKM tracking file built from publicly available Illumina Human BodyMap 2.0 data brain.fpkm_tracking.gz with the format:
  

  tracking_id	class_code	nearest_ref_id	gene_id	gene_short_name	tss_id	locus	length	coverage	FPKM	FPKM_conf_lo	FPKM_conf_hi	FPKM_status
  ENST00000368372	-	-	ENSG00000229631	AL603926.1	TSS84330	GL000223.1:149786-180454	1789	6.76132	0.67731	0.55584	0.798781	OK
  ...	...	...	...	...	...	...	...	...	...	...	...	...

  
  
• GENCODE GTF (transcript/gene) files
  GENCODE GTF (transcript/gene) files from ENCODE project are accepted.
  Example:
  • RNA-seq from ENCODE/Caltech (filter to genes/transcript GENCODE V3c)
  • Long RNA-seq from ENCODE/Cold Spring Harbor Lab (filter to Transcript Ensembl V65).
  • mouse_brain.gtf.gz from ENCODE with the format:
    

  chr9	ENSEMBL65	transcript	65952282	65953531	0.030601	-	.	gene_id "ENSMUSG00000032382"; transcript_id "ENSMUST00000146893"; RPKM1 "0.023032"; RPKM2 "0.078775"; iIDR "0.072";
  chr2	ENSEMBL65	transcript	32218530	32223048	1.581027	+	.	gene_id "ENSMUSG00000039205"; transcript_id "ENSMUST00000132028"; RPKM1 "1.463149"; RPKM2 "3.796865"; iIDR "0.009";
  ...	...	...	...	...	...	...	...	...

  
  
• Kallisto (transcript/gene) files
  PPIXpress uses the reported TPM value to quantify abundance.
  
  
• RSEM (transcript/gene) files
  Example: data from BLUEPRINT project (BLUEPRINT DCC PORTAL). Here, PPIXpress also uses the reported TPM value (rather than FPKM) to quantify the abundance.
  
  
• TCGA (RNA-seq V2) (isoform/gene) RSEM files
  Normalized RNA-seq V2 files from TCGA are also a suitable expression input for the tool.
  The files are assumed to hold TCGA expression data if the associated headers and identifier descriptions are found.
  Specifically, gene expression data from TCGA should have the header:

  gene_id	normalized_count
  A1BG|1	108.8347
  ...	...

  and transcript expression data from TCGA should have the header:

  isoform_id	normalized_count
  uc010uoa.1	0.0000
  ...	...

  Note:
  - Gene annotations are converted from HGNC identifiers to Ensembl genes using the genenames.org-webservice
  - Transcripts are converted from UCSC transcripts to Ensembl transcripts using UCSC’s hg19 MySQL (including all updates to transcript versions).
  - To ensure the best mapping possible, Ensembl data for assembly GRCh37 is automatically enforced.
  - If Ensembl identifiers are found instead, a plain expression file (see below) with a header is assumed by PPIXpress.
  
  Example: BRCA_tumor.normalized RSEM.gz is a BRCA tumor sample taken from TCGA.
  
  
• Plain transcript/gene expression text files
  Expression data can also be supplied line-wise in a text file of the following format with arbitrary whitespaces:

  ENST00000584459	123.456
  ENST00000451927	42
  ENST00000582791	0.0000021
  ...	...

  Note:
  - Genes/transcript are assumed to be given as Ensembl identifiers and a header is not necessary.
  - In the best case, every gene/transcript should only be reported once in the file.
  - If the user’s original data contains multiple probes for the same gene/transcript, for example, he/she can decide in prior how those should be used (average, max,... ).
  - However, if an identifier occurs multiple times throughout the file, only the last expression value is assigned to the gene/transcript and no processing takes place.
  

PPIXpress run options

• Protein Interaction Data Options
  
  • Add STRING weights: Adds weights to the original network using STRING v10 (STRING weights are probabilities of functional association). Interactions that are not in STRING are removed from the network.
    
  • Update UniProt accessions: Update outdated UniProt accessions to their current primary accessions.
    
  • Only local DDI data: Only use local domain interaction data, thus current 3did data is not retrieved and only the precompiled set from DOMINE/IDDI/iPfam is used.
    
  • Include ELM data: Retrieve and integrate ELM motifs and interactions. The retrieval of ELM data might increase PPIXpress runtime significantly and will slow down network visualization.
    
• Processed Expression Data Options
  
  • Gene-level only: Enforces gene-level processing even if transcript expression data is given.
    
  • Normalize transcripts: Normalizes abundance of transcripts by the length of the transcript (only for major transcript file, applied after threshold, useful for count data).
    
  • Expression level:
    • Use threshold: Changes the expression-level threshold to [threshold] (default).
      Threshold is a positive number that defines the minimum expression value for a transcript to be considered as expressed.
      The higher the threshold is, the fewer significant protein interactions will be found.
      The default threshold is set to 1.0 (recommended)
      
    • Use percentage: Changes the expression-level threshold to the [percentile]-th percentile of the expression data.
      The default percentile is set to 0.
      
• Run Options
  
  • Remove decay transcripts: Proteins whose coding transcript is subject to degradation are not removed from the constructed network (default: remove if tagged as ’nonsense-mediated decay’ or ’non-stop decay’).
    
  • Output reference network: Report the according unpruned reference protein interaction network (and optionally the domain interaction network and the longest isoforms assumed) from the mapping stage of the method.
    
  • Output DDINs: Enables the output of underlying condition-specific domain-domain interaction network(s).
    
  • Output major transcripts: Enables the output of the major transcript per protein (including its abundance).
    
  • Report gene abundance: Outputs major transcript per protein but reports the abundance as the sum of all its expressed coding transcripts. Major transcripts will also be included in outputs
    
• Use PPICompare-required options: Include options required for PPICompare (Output DDINs and Output major transcripts).

Examples for run set-up

Sample summary file Sample summary file can be viewed and downloaded from the window on the right side of Running Progress tab once the PPIXpress run is finished. It gives an overview of the condition-specific protein networks built using each expression data.

Sample	Matched output	#Proteins	#Interactions	Threshold
expression_1.txt	1_ppin.txt	5770	69923	1.0
expression_2.txt	2_ppin.txt	3983	39304	1.0

• Sample: Name of the input expression data file
• Matched output: Name of the output file containing the protein network specifically build using each expression data (Condition-specific protein network)
• #Proteins: Number of proteins in the condition-specific protein network
• #Interactions: Number of interactions in the condition-specific protein network
• Threshold: Expression threshold used to filter the transcripts/genes expression

Pipeline log file Once the PPIXpress run is finished, select Download Log File on the left side of Running Progress tab to download the log containing run configurations.

Main output file Select Download Result Files to download main output files, including:

• *_ppin.txt: Condition-specific protein interaction network.
• *_ddin.txt: Condition-specific domain-domain interaction network built using expression_1.txt
• *_major-transcripts.txt: Major transcript per protein in the condition-specific protein interaction network built using expression_1.txt - requires Output major transcripts
Note:
The asterisk (*) is a number (1, 2, 3,...) indicating the expression data which the reference network uses to prune.
This number corresponds to the Sample and Matched output field in Sample summary file
For example: 1_ppin.txt, 1_ddin.txt and 1_major-transcripts.txt are the condition-specific protein interaction network, domain-domain interaction network and list of major transcript per protein built using expression_1.txt.


• reference_ppin.txt: Unpruned reference protein interaction network - requires Output reference network
• reference_ddin.txt: Unpruned reference domain-domain interaction network - requires Output reference network and Output DDINs
• reference_major-transcripts.txt: Major transcript per protein in the unpruned reference protein interaction network - requires Output reference network and Output major transcripts

Build differential network (PPICompare) This option is only available if users run the analysis with self-uploaded data and select PPICompare-required options (Output DDINs and Output major transcripts).

PPICompare compare at least two PPIXpress-constructed PPINs belonging to two different sample types/conditions/contexts to build the differential network. Therefore, users should upload at least two expression data files that represent two conditions to compare.

Users can assign the PPINs to each sample group and forward those to PPICompare to build the differential network between two groups. The PPINs marked with "Not selected" will not be included to PPICompare analysis.

This dialog is shown when users select Build differential network on the left side of Running Progress tab.

In this example, the PPINs pruned using the transcription data for Bulk B cells (4 samples) and CD8 T cells (4 samples) will be forwarded to PPICompare. The differential network between B cells and T cells will be constructed in PPICompare.

Subnetwork visualization Select Visualize Condition-Specific Networks to open Network Visualization tab where user can select a protein and inspect the condition-specific network which the protein involves in.
Once the PPI network is rendered and displayed, the queried protein will be shown in the center. User can toggle between network layouts, customize the PPI network and expand the protein node to see its domains and DDI network. (Domains and DDI network display requires Output DDINs).

Download The PPIXpress standalone version is available on SourceForge.

Documentation The documentation to PPIXpress standalone version is available on SourceForge.

Citation Stand-alone software version: Will, T., & Helms, V. (2016). PPIXpress: construction of condition-specific protein interaction networks based on transcript expression. Bioinformatics, 32(4), 571-578. [Link to paper]
Webserver version: Do, H.T.T., & Helms, V. (2024) [Link to paper]

Contact Webserver support: dhttrang[at]bioinformatik.uni-saarland.de
Corresponding author: volkhard.helms[at]bioinformatik.uni-saarland.de

Icons and animated icons created by Freepik - Flaticon: 1 2