Plot

Volcano plot - all experiments

About

Individual volcano plots for each target with results from all RHAPSODY experiments plotted. 

• Shape: Points are shaped according to omics type. 
• Color: Points with an adjusted p-value <0.05 are shown in blue, while non-significant points are shown in gray. 

Empty plots indicates that the specific target were not measured in RHAPSODY experiments. 

Plot

Volcano plot - Best results from each target

Table

About

Omics-specific volcano plots with the best result for each tissue based on lowest adjusted p-value. Best results (lowest adjusted p-value) for all targets is shown as background points. 

• Shape: Points are shaped according to omics type. 
• Color: Points with an adjusted p-value <0.05 are shown in blue, while non-significant points are shown in gray. 
• Background: The best results for all targets detected in RHAPSODY experiments are included as a reference background in light gray with a reduced size. 

The best results from each target is shown in a table format together with the experiment-specific information, including:

• Experiment: Experimental setting within the RHAPSODY consortium. Find more information on the experiments here. 
• Tissue: The tissue the target was detected in. The tissue availability is omics dependent: 
  • Adipose tissue: Transcriptomics. 
  • Blood: Proteomics. 
  • Islets: Transcriptomics. 
  • Islets LCM (laser capture microdissected): Transcriptomics. 
  • Liver: Transcriptomics. 
  • Muscles: Transcriptomics. 
• Comparison: Experimental setup or modeling. 
• log10 fold change: log10-transformed fold change (FC). 
• Adj. p-value: FDR adjusted p-value. 

Significant results with an adjusted p-value <0.05 are shown in blue, while non-significant points are shown in gray. 

Plot

Disease association in scientific articles

About

Text mining results for target-disease (any) association based on in-house and PHAROS data.

• In-house text-mining  
  • UniProt AC - all diseases co-mentions (PubMed): Sum of diabetes and all other diseases co-mentions in PubMed abstracts. 
  • UniProt AC - all diseases co-mentions (full-text): Sum of diabetes and all other diseases co-mentions in 15 mio full-text articles. 
• PHAROS  
  • Jensen Score: Gene-disease association score based on biomedical abstracts, taking co-occurrence within and between sentences into account. Absolute counts are reported here. For more information, see Pletscher-Frankild et al, 2015. 
  • PubTator Score: Online tool for curation of PubMed articles. Biological entities (chemicals, diseases, genes, mutations, and species) are annotated using entity-specific text-mining tools. Normalized scores used for ranking in the tool is reported here. For more information, see the website and Wei, 2013. 

To ease the comparison between the proteins, a geometric mean has been calculated for the five parameters. Geometric means can be calculated with the formula: \(GM = \sqrt[n]{x_{1} \cdot x_{2} \cdot ... \cdot x_{n}}\).  Using the geometric mean instead of the arithmetic mean (normally known as mean or average) to summarize our data is beneficial for two reasons:

• The geometric mean is not affected by the scale of the data. We can therefore summarize our data without normalizing it. 
• The geometric mean is less sensitive to outliers than the arithmetic mean, but still more sensitive than the median. 
• The geometric mean is also more stable than the median, which can change dramatically when adding or removing just a few input values if your sample size is small.  

However, it is important to note that the geometric mean does not handle zeros and negative values. It is therefore not useful for all data types. 

Plot

Disease co-mentions

About

The in-house text-mining has been conducted in two modes for PubMed abstracts and full-text papers: 

• UniProt AC - all diseases except diabetes co-mentions 
• UniProt AC - diabetes co-mentions 

Using these text-mining results, we can compare the number of co-mentions associated with diabetes to all other diseases. 

The plot is split into three panel (left-to-right): 

• A proportional area plot (left panel) shows the percentage of co-mentions in full text articles that are diabetes-related.  
• The first bar plot (middle panel) shows the co-mentions count for diabetes in dark blue and the co-mentions count for all diseases but diabetes in light blue derived from 15 million full-text articles (in-house resource, see Westergaard et al. PLoS Computational Biology 2018). 
• The second bar plot (right panel) shows similar to the middle panel co-mentions count for diabetes in dark blue and the co-mentions count for all diseases but diabetes in light blue. The data is obtained from 15 million abstracts (in-house resource, see Westergaard et al. PLoS Computational Biology 2018). 

The plot has been sorted according to highest full-text all diabetes co-mentions. 

All interactions

PPI networks

High-confidence interactions only

PPI networks

Summary table

About

Protein-protein interaction networks based on physical links data from the STRING database. See Szklarczyk et al, 2018 for a description of the resource.

Network features:

• Interaction partners (nodes) are colored according to the confidence score of their interaction with the primary targets.
  • High confidence interaction partners (confidence score > 0.7) are shown in blue
  • Suggested interaction partners (confidence score ≤ 0.7) are shown in gray
• Interactions (edges) are shown according the confidence score of the interaction
  • High confidence interactions (confidence score > 0.7) are shown as solid line (—)
  • Suggested interactions (confidence score ≤ 0.7) are show as dashed lines (- - -)

Summary table:

• Mean diabetes textmining %: The avarage diabetes percentage of diabetes co-mentions to total co-mentions across the network. Based on textmining from Westergaard et al
• Associated with diabetes: Number of nodes that have been associated with diabetes mellitus (Open Targets)
• Suggested diabetes biomarkers: Number of nodes that have been identified as potential biomarkers for Type 2 diabetes mellitus
• Potential interaction partners: Number of interaction partners below the confidence threshold (confidence score ≤ 0.7)
• High-confidence interaction partners: Number of interaction partners above the confidence threshold (confidence score > 0.7)
• Total interaction partners: Total number of interaction partners

In cases where none of the targets have any interaction partners (all or high-confident), the following error messages will be shown:  

• All interactions: Unfortunately, no interactions were found for any of the selected targets.
• High-confidence interactions: Unfortunately, no high-confidence interactions were found for any of the selected targets.

All interactions