This is a step by step tutorial for the tools in ADEPTUS. First,from the menu you can access the tutorial, main, and download pages at any time. You can also contact us, and our team at ACGT.

ADEPTUS offers three types of analysis:

Analyze a gene list

This page allows different enrichments on an input gene list as gene symbols. You can also present the gene list as a network. The gene list input can be given by filling the box or by upload a “.txt” file. The genes can be separate by space, comma, tab or new-line (“Enter”).

 Gene Ontology Enrichment – we use TANGO, a tool for analysis of the GO hierarchy, see Expander.

 Pathway Enrichment – we use Fisher’s exact test with FDR correction. For example, when we run the enrichment on this list {TP53, BRCA1, BRCA2}, the result table:

The table is ordered by the q-values, from low (most significant) to high.

 Disease ontology enrichment – In this enrichment we test the gene list against the diseases that pass our statistical analyses. The test checks if the gene list has unusual ranks in our pre-computed gene scores for each disease. The analysis is similar to GSEA but the input is reversed: the gene ranking is kept fixed and the user provides the gene list.

 View gene network – This option will present a network based on the genes as nodes, and the edges between (Genemania PPI by default). More about the graph page in the Analyze a disease or tissue part.

All the enrichment results can be downloaded as a txt summary table:

Analyze a disease or tissue

This option will lead you to tree-like graph (DAG) of our well classified diseases labels. Each disease is represented by a node, and the edges represent “is a” relations. This graph is active:

• By left mouse-click on the graph area you can budge the whole graph.
• By left mouse-click on node you can grab him.
• Zoom-in - By scroll the middle mouse button on the graph area.

Example - click on the graph area

At the left side of the page there is a bar with “choose” options:

Choosing the first label – “A control label of tissue” (1) will open new window below with list of tissues, by choosing one of them(2) and pressing on the “View Gene Network”(3) button beneath it, you will get a network visualization of the genes in that control term.

The second option – “Tissue-based Disease Ontology terms” will also open a new window with a list of tissues below. This time, selecting a tissue and clicking on the button will produce a visualization of the relevant disease labels:

This tree contains all the well classified disease labels for the selected tissue. In the label network, left mouse-click on disease node will pop up a balloon with “GO!” button for moving to the label’s gene network.

Well classified diseases for breast tissue

By choosing the third option – “General Disease Ontology terms” and click on the “Show Tree” button we also will get new tree at the right side of the page, this tree contain the non-tissue well classified diseases. At this tree, again, left mouse-click on disease node will pop up the “GO!” button balloon, and by clicking on him we will get the relevant network graph.

Graph page

This page is the result of each disease query. Also, we can get into this page from “Analyze a gene list” → “View Gene Network” for user given gene list. The graph page contains many options:

1. The graph area – similar to the diseases tree – by left mouse-click on the graph area you can move the whole graph, and by left mouse-click on gene node you can move the specific node. Although, in this graph, by clicking on the node we get additional information:
   
   • Links to other databases:
     

     

     
     Notice that genes with “-“ in their official gene symbols will have “_” instead in our graph presentation (e.g., “NKX3-1” → “NKX3_1”).

   • Gene report: a meta-analysis on the gene’s information in the database. Here a new bar will slide into the right side of the page. The bar title will show the gene and the label (e.g., “Report for gene NISCH in control; prostate”). After that, there are two tables. The first one includes the gene’s scores of our univariate analysis (e.g., ROC scores). The second table gives the statistics of the label in the database. Finally, we give a bar plot that shows the -log10 p-values of the specific gene in all datasets that were used in the label’s analysis. This table can be used for meta-analysis or replicability analysis for the query gene.

     

     
     
   • Node visualization: The coloring of gene nodes is separated into 4 quarters, each one has his color options which signifies information from different data sources:

     

     
     
   • Number of nodes in a label’s network: The well classified label networks will contain at least 10 nodes and at most 200. If there are less than 10 nodes that pass the statistical analyses then the network will show the 10 nodes with the highest PN-ROC scores.
     

2. Graph control:

   

3. Graph visualization:

   

   
   Node selection can be done in two ways – (1) select node-by-node by mouse-click & pressing “Ctrl” key together. (2) Press “Ctrl” key and drag the mouse with continues click over the selected nodes.

4. Layout option: by choosing this menu 6 more buttons will slide from the left side. Each button is network layout option, and by clicking on each one the network visualization will change –

   

   
   
5. Data & analysis: This menu will open at the left side too. In this menu users can manipulate the data by setting different thresholds for gene scores. Pressing “Update” gives the new network. The default gene score thresholds are: PN-ROC (intensity) 0.7, specificity ROC 0.7, and replicability 0.5.

   

   At the top there is an option to select the edges view, each option represent interactions.
   

Brief terminology of the gene scores (see the paper for more details):

1. PN- and PB-ROC scores – these are based on the separation of the expression levels in the label as compared to direct and indirect control samples in the database. The coloring of the upper side of the nodes tells if the gene is up-regulated (green) or down-regulated (red).


2. Specificity score – this score is based on the separation of the expression levels in the label as compared to similar diseases.


3. Replicability score – this score is based on the p-values of the separation between the label’s samples and their controls within the datasets.

Analyze a profile

There are two different options.

• Predict cancer site from mutated genes – the user can insert few different samples, each sample is gene list, separated by spaces, and samples separated by newline. The gene list of each sample represents the set of somatically mutated genes in the cancer sample. The classifiers are based on REF.

  The result table shows the predictions as probabilities for association between cancer subtypes (rows) and samples (columns). Here is an example:

  



• Predict phenotype from expression – here the user uploads a matrix of gene expression. See this file for example. The output is a table in a similar format as above. However, the predictions are for association between the samples and our well-classified labels.