The Connectivity Map web-site and database provides access to over 6000 experiments where approximately 1300 individual drugs, many of them food and drug administration (FDA) approved, are applied to human cancer cell-lines while gene expression is measured with Affymetrix cDNA microarrays after six hours. The Connectivity Map web-site provides a querying tool for prioritizing drugs based on matched gene expression data provided by users. Here we implemented a Java program called Drug Pair Seeker (DPS) that attempts to predict and prioritize, instead of single drugs, pairs of drugs using the same dataset. Users can enter lists of up and down differentially expressed genes from their own experiments to receive a ranked list of drug combinations that would either reverse or aggravate the condition of their cells or tissue using a simple formula.

The installer for DPS is available for download at: http://www.maayanlab.net/DPS. The installer is compatible with Windows, Mac OS X, and Linux.

DPS requires that you have at least Java 7 installed. You can get the latest version of Java at: http://www.java.com/getjava.

The installer will install the required files to run DPS, shortcuts to run it, and optionally the source code and documentation for DPS.

To perform a simple analysis, import your gene list containing up-regulated and down-regulated genes by clicking on the "Overexpressed Genes" and "Underexpressed Genes" buttons respectively.

Specify an output location. If you don't specify an output location, the default output location is where the DPS program exists.

Using the radio buttons, you can select to either find drugs that reverse the condition that is described by your pair of gene lists or aggravate that condition. Use the drop down list to select the drug data source you wish to use. The options are New L1000 CMAP and Old AFFY CMAP. Clicking on "Find Drug Pairs" starts the search.

Once the progress bar completes, you may click "View Results" to view the results in your browser.

The HTML output allows you to collapse parts of the output so the results do not overwhelm you. Clicking on the bar opens further information about the pair or the genes that they affect. The output lists the pairs of drugs that best reverse/aggravate the condition and are ranked based on their performance. The rank of each individual drug if it were to be used alone is also given in parenthesis. Performance of a drug is evaluated based on score = coverage - conflict. Coverage refers to the number of desirable targets that the drug affects. In the case of reversing a condition, the desirable targets are the genes described by the condition that the drug reverses. Conflict refers to the number undesirable targets that the drug affects. In the case of reversing a condition, the undesirable targets are the genes described by the condition that the drug further affects in the same undesired direction.

If you already know which drug you would like to pair with other drugs, you can restrict the analysis to focus on the specific drug and find other drugs that would best pair with it for the best coverage. Using the drop down menu, you can select the drug that you would like to restrict the search to focus on.

Enter the command shown below to run DPS in batch mode. By default, the program will search for drugs that aggravate the condition described by the input gene lists, but with the "-r" switch, it will reverse the condition. data_source is a switch that can either be set to "New", meaning the new L1000 CMAP data or "Old", meaning the old AFFY CMAP. It is important to increase the java heap size to at least 1gb.

If we run DPS using these sample up and down gene lists, we obtain these sample results.

Supporting tables for the article "Renal-protective effect for ACEI and HDACI inferred from CMAP" currently under review: Table S1, Table S2, Table S3, Table S4, Table S5, Table S6, Table S7, Table S8.