Survey sequencing of mRNA gene products provides an indirect means of generating gene expression fingerprints for cancer cells and their normal counterparts. Digital Differential Display (DDD) is a computer method for comparing these fingerprints. Using a statistical test, genes whose expression levels differ significantly from one tissue to the next are identified and shown to the user.

Compare organs of the body

By analyzing differential gene expression patterns, it may be possible to identify genes that contribute to a cell's unique characteristics. For example, those which make a muscle cell different from a skin cell or a nerve cell.

Compare stages of prostate cancer

Along similar lines, DDD might be used to identify genes whose expression levels differ between normal, premalignant, and cancerous tissues. A few of these genes may actually play a role in the process of cancer development. Others may be a consequence of cancer rather than a cause, but may still be useful as markers for the early detection of cancer. Still others might be used to predict responsiveness to different cancer therapies.

DDD Basics

The foundation of DDD is UniGene. UniGene employs a conservative method to assign all the human EST sequences that meet minimal standards of quality to distinct "clusters", each representing a unique human expressed gene. DDD takes advantage of UniGene by comparing the number of times sequences from different libraries were assigned to a particular UniGene cluster. This has the advantage that DDD will only report on sequences that we have confidence represent bona fide human expressed genes.

There will of course be many differences in the number of sequences contained in each library that are assigned to a particular UniGene cluster, but only some of these differences are likely to reflect biological reality. Therefore DDD employs a statistical method of comparison - The Fisher Exact Test - to identify only those differences that are likely to be real. One important factor in determining statistical relevance is the absolute number of sequences in each library that have been successfully assigned to a UniGene cluster. In many cases there are not enough sequences in dbEST libraries to meet the threshold of significance employed in the Fisher Exact Test. Since DDD will only yield a report if there are differences that exceed this threshold, it is expected that many comparisons will yield nothing.

How to:

1. As with all DDD "pages", a table appears at the top summarizing your choices for comparison. The unit of DDD is the "pool": a pool is a collection of libraries (minimum = 1) that you wanh to group together on the basis of your own criteria for comparison with other pools you will define. The yellow background indicates the pool you are currently editing. So, on the initial form you see the following:
   
2. Below that, you see a text box inviting you to name the pool of sequences that you are about to collect. To name the current (initial) pool, enter text in the box: 
3. To select the library, or libraries, to include in the pool, scroll down and using your mouse check the box to the left of each library you want to include. Libraries that have been "normalized" are indicated by N, whereas those that have been subtracted are indicated by S. Note that each library has a numerical "ID": clicking on the ID will link you to a summary report of that library: 
4. Now, Click the button labeled "Accept Changes".
5. Because you must have a minimum of two pools to compare to each other, you will now repeat the process. Note that the table at the top of the page indicates the numerical library "id" of those you chose to include in Pool A, as well as the name of the pool. The column labeled "ESTs" indicates the number of sequences from the Pool that have been assigned to a UniGene cluster. The yellow background now indicates that it is Pool B that you are editing. 
   
6. Repeat the previous steps.
7. Upon clicking "Accept Changes" this second time, DDD goes on to compare the two pools. For each sequence in a pool that has been mapped to a UniGene cluster, DDD will compare the number of sequences from that cluster that appears in every other pool. DDD will report when there is a significant difference between the number of sequences that belong to a cluster in one pool versus another pool.
   
   NOTE: DDD only reports on a cluster if at least one pairwise comparison results in a statistically significant difference. However, when such a difference is found, the DDD report includes information on the number of times that UniGene cluster was represented in all pools, although the report distinguishes those differences that are statistically significant. You may find it useful to include a "control" pool that contains a large number of sequences that because of library source is unlikely to share sequences with your libraries of interest. Since there will be many statistically significant differences between the control and other pools, DDD will report on many more genes, and while many of the differences between your libraries of interest will not be statistically significant, you may find them biologically interesting nevertheless.
8. To add another pool, click on "NEW" and repeat the above process.
9. To change one of your previous choices, click on "EDIT".

DDD Output:

The description of the pools will remain at the top of the DDD page, whereas the results will be found below.

The DDD results are delivered in a table (below). Columns labeled with letters represent their respective pool (i.e., column "A" = "Pool A") and will contain at least part of the name. Each row represents a UniGene cluster: the column marked "Gene Description" contains the name of the cluster, whereas "Gene Index" refers to the numerical "ID" of the cluster. Clicking on this "ID" will link you to a summary report for the UniGene cluster.

The DDD results table provides several pieces of information for each gene (see figure below). The numerical value (A) is the fraction of sequences within the pool that mapped to the cluster shown. The dot (B) is merely a visual aid that reflects the numerical values. If any pool participates in a statistically significant pairwise comparison with another pool, the relationship is indicated (C). For example, in the table sample below, in the first row, first column, "A < E" indicates that the greater amount of sequences found in Pool E versus Pool A for the gene Hemoglobin gamma-G is statistically significant. Likewise, in the same row, column "E" lists the same relationship from the point of view of Pool E: "E > A". Note that even though Hemoglobin gamma-G cDNAs are not found in any of the four prostate samples (Pool A - Pool D), nevertheless, the pairwise comparison between the "control" (Pool E) and Pool D is statistically significant, while the pairwise comparison of Pool C with Pool E is not statistically significant.

You can get started by following the DDD link in the organism-specific section of the sidebar of any UniGene Web page (set to Homo sapiens for this info page). Your browser may take more than a few seconds to display the resulting data - please be patient!