100% Spot Matching: Complete Expression Profiles with Delta2D

Quick Summary
Comparison to the Traditional Spot Matching Approach
How Delta2D Achieves 100% Spot Matching

Quick Summary

Delta2D's 100% spot matching produces complete expression profiles for every protein. Besides higher throughput, this leads to significantly improved statistical confidence, so you can, for example, identify more biomarker candidates from the same experiment. Other approaches to spot detection and matching lead to inconsistencies like missing values in expression profiles, and ambiguities in the profiles themselves. Delta2D's 100% spot matching is based on advanced image processing methods. It was introduced by DECODON in 2003.

Here is a set of gels analyzed with Delta2D's 100% spot matching:

Note: The fourth image (with red spot boundaries) is a synthetic fusion image that is used in the process, see below.

You see that the spot boundary patterns on each image are essentially the same. When a spot is not present on one of the gel images, Delta2D will still produce a segment there, with a quantity near zero. Notice that the expression profiles in the table have no gaps. This is comparable to the situation in transcriptomics: in a set of DNA arrays, you can always find corresponding spots across multiple arrays — they are at the same row and column on each array.

The Benefits of Delta2D's 100% Spot Matching

Better statistical analysis

Complete expression profiles mean that more changes in spot patterns can be identified with statistical significance. Thus you can, for example, identify a larger set of biomarker candidates from the same raw data.

Increased productivity

Spot detection and editing is only done for the fusion image, as opposed to each image. For a 24 gel experiment this means you only have to check and correct one set of spot boundaries, as opposed to 24 different sets.

Better reproducibility

No hand-editing of spot matches means less potential for human error and operator-dependent variation. Furthermore, variation that is due to variations in the laboratory process, like sample preparation, or electrophoresis conditions etc. are much easier to detect when they are not obscured by matching problems.

Comparison to the Traditional Spot Matching Approach

With traditional software packages, you have to edit the detected spots on each gel image, and resolve matching conflicts by inspecting all images involved. Here is an example of spot patterns and  table of spot quantities produced by the traditional "separate detection on each gel image" approach (default parameters used, no spot editing).

While you can correct many of the mismatches by hand, this process takes a lot of time and still does not resolve all ambiguities. The remaining spot matching ambiguities are resolved by the traditional software according to some black-box rules, producing frequent "missing values" in the expression profiles. A missing value can mean that there really was no spot on that gel, that there was a spot, but the software did not detect it, or that the software was unable to find the correct matching. In any case the expression profiles contain gaps that diminish the statistical confidence in subsequent quantitative analyses.*

How Delta2D Achieves 100% Spot Matching

Step 1: Gel image warping

Gel image warping compensates for running differences between different gels. After warping, corresponding spots will have the same position on every image. Here is a set of four gel images made from different samples:

As you can see, the spot patterns are quite different. One image is selected as a reference image (blue), the other images are warped onto it. Here this process is illustrated with dual channel images where the reference gel is shown in blue. Before warping, the spot patterns do not match due to running differences between gels:

After warping, corresponding spots have the same position:

Now Delta2D knows how to compensate running differences when processing images.

Step 2: Image fusion

Image fusion means to combine multiple image into a new, synthetic image. We use a modified average of the values of corresponding pixels across all images. Details are describend in the paper: Luhn S, Berth M, Hecker M, Bernhardt J: Using standard positions and image fusion to create proteome maps from collections of two-dimensional gel electrophoresis images. PROTEOMICS 2003 Jul;3(7):1117-27. Digital Object Identifier (DOI): 10.1002/pmic.200300433 PubMed Entry: PMID: 12872213.

Even though the fused image is a synthetic one, it looks like a real gel image, and, more importantly, all spots from the experiment are represented on it:

We call this a proteome map because it shows all spots on one unified image.

Step 3: Spot detection on fused image

Spot detection on the fused image produces a consensus spot pattern:

Spot detection can be corrected here, once per project as opposed to once per gel in the traditional approach.

Step 4: Transfer Spots

The consensus spot pattern is then transferred to the original images. Since Delta2D knows about the running differences between gels, it can transfer spot boundaries from the fused image to the corresponding regions on the original images.

Spot shapes are made wider or narrower depending on the spots present on the target images, and then quantitated. When a protein is not present on a particular gel, the corresponding region will have a quantity that is near zero. As all spot patterns originate from the consensus pattern on the fused image, Delta2D can achieve 100% spot matching, giving a complete expression profile for every protein.