Usage¶

There are two main options available for runing DeCoaD. The reuired input and the output of the program depend on the selected option, as explained below.

Option 1: Show diseases similar to a particular disease¶

If this option is chosen no other input is required on the main DeCoaD page. Hitting CONTINUE takes the user to another page to enter the required input parameters as described below.

Input

The input form has six fields divided into three parts. For each field, a help content can be accessed by clicking on the question mark next to it. In the first part the user provides the main input, i.e. the ID of the disease of interest, which may or may not be present in the network. The list of the included diseases can be accessed by clicking on the link in the provided help content. If the input disease is not in the list, the user also needs to provide the IDs of the genes associated with the disease in the appropriate text box. The listed genes, however, must be present in the network. The list of included genes can be seen by clicking on the provided link in the help content. Even if the disease is already present in the network, the user may enter a list of associated genes. In this case, however, the existing gene associations of the disease are ignored, unless they are entered in the text box. Such changes in the network, requires all the weight vectors to be recalculated, which results in an increase in the running time of the program. The remaining four input fields, which are pre-populated with the default values, are used to limit the number of reported similar diseases and clusters. The user may set either the maximum number of reported diseases/clusters (limit by rank) or the cutoffs for correlation and probability to be considered significant (limit by minimum correlation/probability). If the second option (setting the cutoff) is used and if the cutoff is so high that no significant disease/cluster is found, the disease/cluster with the highest similarity/probability is reported. After filling out the input form one may run the program by hitting the RUN button at the bottom of the page. The other two buttons, RESET and EXAMPLE, are used to reset the form and to pre-populate the fields with an example respectively. It is worth noting that the disease network required by DeCoaD is loaded when the program is run, and so there is no need to be provided as an input by the user. The network is periodically updated to reflect changes in the databases that it is based on.

Output

The result page is divided into two sections. The first section summarizes the results in three subsections: “graphical summary”, “similar diseases” and “clusters containing the disease”. In the graphical summary subsection the CTD disease database is used to show a directed graph whose leaves, colored in blue, are the disease of interest and the top ranking similar diseases. The darker shades of blue are indicative of higher correlations between the identified diseases and the input disease. The graph also shows the hierarchical structure of the disease families containing these diseases. Each node (disease) in this graph is linked to its description in the CTD database. In the next subsection of the first part of the result page, the names of the top-ranking diseases and their correlations with the input disease are given. The list of the cluster IDs containing the input disease and the corresponding membership probabilities are tabulated in the third subsection. Each cluster ID is linked to a web page that lists, in descending order, the membership probabilities of all diseases. It should be noted that, as mentioned before, when new gene associations are provided in the input page, the weights and probabilities have to be recalculated. To speed up this process, the probabilities are calculated approximately. In such cases another column, which gives an upper bound for the error caused by the approximation, is added to the table.

The second section of the output page provides an interface to SaddleSum. The user has the option to perform enrichment analysis for the input disease itself or for any cluster that contains it. For a description of Saddlesum and its required input parameters see its documentation page.

Option 2: Show a list of significantly related diseases¶

If (on the main DeCoaD page) the second option is chosen, a cutoff to limit the number of listed disease pairs must be entered as input. The number of pairs shown can be limited by either a similarity cutoff or a rank cutoff. Each of these two options is explained by a help content that can be accessed by clicking on the question mark next to it. Hitting CONTINUE opens a new page that shows the output, i.e. the list of all disease pairs that clear the cutoff. For each disease pair listed, the corresponding similarity is also reported. Additionally, a mailto link (Submit evidence) is provided so the users who are aware of publications confirming the relation between the two diseases can email us the PMIDs of these publications.