Aranet - Tutorial

Network-assisted prediction method by AraNet version 2

1. Find new members of a pathway allows users to predict new candidate genes for a phenotype or biological pathways. Functionally associated genes are wired to another in the network. Thus, genes that are highly connected to genes of known function (user submitted query genes : genes that are known to participate in certain biological pathways or that lead to certain phenotypic outcome) are candidate genes of the pathway. This method provides top 100 candidate genes.

1. Submit your query genes

Click “Find new members of a pathway” on Network-search page.

(1) Species selection box. Select your species by clicking this box. You can find 28 plant species available including Arabidopsis thaliana

(2) Insert your query genes in the text box. The gene names must be separated by tab, comma, white space or new line. A maximum of 300 genes are available. It is important to match the gene name form to the example gene names provided in the table located in the lower part of this page.

(3) Or you could select our example query genes for a test run.

Click ‘submit’ button to start the analysis. This may take about 30~60 seconds.

2. Interpreting the results

(1) Table of genes with Arabidopsis thaliana orthologs in AraNet. This table does not show when you run your analysis with Arabidopsis thaliana genes. The rest of the result would be based on these genes.

(2) Table of genes without orthologs or genes that are not in current annotation version which cannot be used for analysis. This table does not show when you run your analysis with Arabidopsis thaliana genes

A. Area Under ROC curve

If the query genes are well connected to one another by the network, other genes that are connected to the queries have a high probability of being a new candidate gene. This assumption makes us possible to estimate the prediction power of the network by measuring inter-connectivity of the genes. Candidate genes are ranked by how well they are connected to the query genes. (calculated connectivity scores) Query genes that are well connected to one another will be highly ranked. Prediction power by this rank order can be assessed by Receiver Operating Characteristc (ROC) curve which represents the true positive rate (TPR) for the given cost of false positive rate (FPR) as described here.

The ROC curve can be summarized in a single Area under ROC curve (AUC) score which is a numerical measure of prediction power.

AUC = 0.5 : random prediction
AUC = 1.0 : perfect prediction

B. Network visualization

The connection of query genes will be visualized by Cytoscape. This requires installation of adobe flash player.

Clicking on (1) node or (2) edge will present you the information.

C. Query gene connections

The query genes that are connected to each other are shown on a table by rank order.

(1) Rank by total connectivity: Guilt-by-association principle prioritizes query genes by their connectivity to other query genes.

(2) Other plant species ortholog gene ID: This column shows gene ID of other plant species that are orthologous to Arabidopsis genes. It won’t be shown if you query in Arabidopsis genes.

(3) Locus name: Gives the name of Arabidopsis query genes or Arabidopsis ortholog genes of other species genes. Symbol represents alternative common names.

(4) Score: Sum of LLS (Log Likelihood Scores) of network links to all other query genes.

(5) Evidences : list of subnetworks (see network download page) supporting network-based prediction and their fraction of contribution

(6) #linked query/ # valid query : The number of linked query genes over the number of total valid query genes.

(7) Linked query : Shows a list of linked query genes.

(8) Arabidopsis thaliana GO terms : Shows GO biological process, cellular component, molecular function of query Arabidopsisgenes or Arabidopsis genes that are orthologous to other plant query genes. (when submitted other plant species gene queries)

Clicking Arabidopsis gene name leads to TAIR website for detailed description of the gene.

Genes that are not connected to any other query genes are also shown (disconnected genes). You can infer function from their neighbors (which would not be query genes) by clicking the submit button below. (For inferring function, see ‘Infer function from neighbors’ tutorial)

D. Candidate gene prediction

Candidate genes (orange ovals) that are connected (grey lines) to query genes (green ovals) are visualized in cytoscape. As in tutorial part B, you can get information of edges or nodes by clicking them

AraNet v2 provides list of candidate genes that are ranked by connectivity of genes to the query gene. Maximum 100 predictions are shown on the web but you can download the full result list by clicking all_predictions_report_text_file. The columns are the same as described in tutorial part C.

Open Tutorial

2. Infer function from network neighbors allows users to infer function of an uncharacterized gene from its network neighbors. Genes with similar function tend to be closely connected in the network, the uncharacterized genes (user submitted genes with unknown function) may have similar function with its neighbors. This method assigns neighbor inferred most relevant 30 GO-BP(Gene Ontology-Biological Process) terms to the uncharacterized genes.

Open Tutorial

- Predictions in other plant species

The functional linkages of A. thaliana genes are projected to other plant species genes for network prediction. All of the analysis runs with projected functional links from AraNet. If you run your analysis with other 27 available plant species genes, you will be able to see A. thaliana ortholog partners throughout the analysis. You may see a number of other plant species having the same A. thaliana ortholog partner. This is because those genes have exactly same orthology score (BLASTp raw score) to the partner gene.


Home	Tutorial	Network-search	Network-download