Searching for concise conclusions on a particular subject within a long list of publications could be difficult and time-consuming, requiring a researcher to read through each publication, identify the sections of interest and analyze the associated figures. In addition, the exponential growth of the genomics field has led to the generation of massive amounts of primary data, often stored in large databases which can be difficult to navigate without bioinformatics expertise. It is also challenging for investigators to determine the significance and strength of a particular result from primary genomics data alone. Therefore, there is a major need for a searchable web interface assembling preliminary results, connecting them with subsequent evidence that reinforces or builds upon the original finding, and conveying the extent to which the results have been functionally validated.
Introduction to the CTD2 Dashboard
The Cancer Target Discovery and Development (CTD2) Network Centers at the Broad Institute, Cold Spring Harbor Laboratories/Memorial Sloan Kettering Cancer Center, and Columbia University have developed the “CTD2 Dashboard”, which compiles the CTD2 Network-generated conclusions, or “observations”, associated with particular subjects (the focus of experiments; they are categorized into classes, such as gene, cell line, animal model, or perturbagen). It provides related evidence (bulk datasets, data-related figures, links to source data, and references) for each observation. Subjects are assigned “roles”, or functions, based on the evidence. CTD2 Center-designated roles are standardized and include biomarkers, disease, master regulators, interactors, oncogenes, perturbagens, candidate drugs, and targets. These roles allow for easier browsing and searching, and help CTD2 Dashboard users draw conclusions about the significance of findings. Importantly, the CTD2 Dashboard was designed to allow easy navigation and use by computational experts, those with little bioinformatics experience, and others in between. Thus, the CTD2 Dashboard is designed for users to quickly and easily find, analyze, and build upon Network-generated experimental results.
The Network has created a classification system to rank observations in the CTD2 Dashboard into “Evidence Tiers” based on the strength of the supporting data. Observations are placed into the appropriate tier, from 1 (lowest) to 3 (highest), based on the extent of characterization associated with the observation. As the Tier level increases, evidence correlation is strengthened and conclusions are more clinically relevant. Briefly, Tier 1 is assigned to preliminary positive observations, Tier 2 is associated with confirmation of primary results in vitro, and Tier 3 labels results that have been validated in a cancer-relevant model in vivo. A major advantage of the Dashboard is that it compiles all of the Network-generated evidence associated with a particular subject in one place, thereby displaying the progression of evidence from Tier 1 to 3. The research community can therefore use the Tier ranking system to determine which observations in the CTD2 Dashboard are robust enough to serve as a basis for further validating results in vivo and developing novel therapeutic targets and biomarkers for cancer.
The CTD2 Dashboard provides users a variety of ways to browse and search observations and make their own discoveries.
To provide an example of CTD2 Dashboard functions, we highlight the many Network-generated observations for the MYC gene. Although MYC is a well-characterized gene, with known roles in cancer, it still remains elusive as a therapeutic target. Future research that builds on existing CTD2 Dashboard observations on MYC could lead to clinically-relevant applications.
Users can browse observations by clicking on a category of subjects on the CTD2 Dashboard homepage.
Figure 1: The Dashboard homepage links to pre-configured browse pages. Separating the observations into categories facilitates easier exploration.
The CTD2 Dashboard provides three browse categories: Biomarkers, Targets, Genes & Proteins; Compounds & Perturbagens; and Disease Context (Figure 1). Clicking on a category will bring the user to a browse page, which displays a table of subjects under that particular category (Figure 2). For each subject, the number of observations in each Tier is listed. A subject is only designated with a role if it has more than one associated observation to support that role. Depending on the evidence available, subjects may have more than one role; users can change the display settings to all roles, or choose particular roles by clicking on the “select roles” button next to the heading on each browse page.
Figure 2: The “Biomarkers, Targets, Genes & Proteins” browse page provides a tabular presentation of class, subject (listed in the “Name” column, for example, MYC), and observations. Clicking on “select roles” will allow the user to change the roles displayed in the browse page table. The browse page is shown here with all possible roles selected (target, biomarker, background, candidate master regulator, interactor, master regulator and oncogene). The MYC gene is designated with the role of “oncogene” based on observations in Tiers 1, 2, and 3.
Users can view all Network-generated observations associated with a particular subject.
Figure 3: Selecting a subject from a browse page table (example shown: MYC) produces a list of associated observations, each with the associated Tier rank and the Network Center that produced the finding. At the top of the page, basic information about the subject is displayed. The subject pages are interactive; users can follow links to the associated entries in Entrez, UniProt and cBioPortal.
From the browse page table, users can click on a specific subject, such as MYC, to see all of the associated observations (Figure 3). Alternatively, users can click on the observation number under a specific Tier, to view only observations in that Tier. Users can then click on the “details” link to navigate to an individual observation page and find the supporting evidence, complete with links to references, data files, figures and/or associated summary stories (Figure 4). Each observation page also contains a table displaying the additional subjects associated with the observation, along with their roles and a link to that subject’s page. This allows users to easily explore all aspects of a particular observation in a single location, rather than perform multiple searches.
Figure 4: Each individual observation page displays the observation, the Tier level, and the associated Center. Below that, a table lists all subjects associated with the observation (such as AKT1 and MCF10A, shown here) and their role in the observation. Each subject is linked to its respective page in the CTD2 Dashboard. The Evidence table lists all of the Center-generated evidence supporting the observation. In the Details column, users can find figures, links to references, data files, and/or associated summary stories.
Users can retrieve evidence and observations pertinent to their queries by searching across subjects using standardized terms and vocabulary.
For example, users can search for genes using synonyms or Entrez Gene ID, proteins with UniProt ID, and compounds with PubChem or Chemical Abstract Service (CAS) numbers. This non-restrictive search feature allows users to customize searches to their individual preferences. The search feature can be found on the homepage. Searching for a subject will retrieve all observations available in the CTD2 Dashboard, regardless of role or Tier.
Users are able to compile a list of genes for further analysis by adding them to the Dashboard’s “Gene Cart” while browsing or searching.
After genes are added to the cart, users can query databases of molecular interactions (DNA-DNA, DNA-protein, and protein-protein) within a variety of tissue- and disease-specific interactomes. By selecting a type and version for interactome viewing, users can create molecular interaction network maps based on the Cellular Networks Knowledge Base (Figure 5). This feature allows users to build interactomes directly from a platform that also has general and experimental details for each gene of interest.
Figure 5: Genes of interest, such as MYC, can be added to the Gene Cart by clicking the + next to the gene. Then, users can click “gene cart” in the menu bar to go to the cart and select a gene of interest to create an interactome. Once a user enters gene cart and clicks “Find gene interactions in Networks”, the CTD2 Dashboard will walk users through steps necessary to produce an interactome.
Dashboard visitors can read stories associated with publications.
Stories are summaries of research findings, written for a broad scientific audience (Figure 6). Stories are written such that users from all research fields can easily grasp the experimental design, analyses and conclusions. When CTD2 Dashboard subjects are mentioned in a story, they are linked to their respective observations pages in the CTD2 Dashboard.
Figure 6: Dashboard stories highlight research from the CTD2 network. Links to the latest stories are displayed on the homepage. Clicking “More stories” will bring the user to a list of all available stories, sorted by date of addition to the CTD2 Dashboard. Users can choose to see observations associated with the story, or view the full story.
A major advantage of the CTD2 Network Dashboard is that it allows users to visualize the flow of research from hypothesized correlations to validated molecular relationships, and to view each finding as an easily digestible summary. The CTD2 Dashboard acts as a “one stop shop,” allowing users to access multiple types of general and experimental information and various external analytical tools from one website. This streamlined process of exploration will hopefully lead to more hypotheses, discoveries and clinical interventions.
As the Dashboard continues to be updated with new Network-generated data, its importance as a resource for the research community will be enhanced. For more information on the CTD2 Dashboard organization including term definitions, please read Navigating and Understanding Dashboard Content. All interested researchers are encouraged to visit the Dashboard and provide feedback.
We would like to acknowledge the CTD2 Dashboard development team: Arman Aksoy, Andrea Califano, Paul Clemons, Vlado Dancik, Tanja Davidsen, Aris Floratis, Daniela Gerhard, Benjamin Gross, Leandro Hermida, Nadia Jaber, Subhashini Jagu, Zhou Ji, Ava Li, Jinyu Li, Jessica Mazerik, Don Monroe, Chris Sander, Stuart Schreiber, Kenneth Smith, and Cliff Wong.