The efficiency and reproducibility of two-dimensional difference gel electrophoresis (2D DIGE) depends on several crucial steps: (i) adequate number of replicate gels, (ii) accurate image acquisition, and (iii) statistically confident protein abundance analysis. The latter is inherently determined by the image analysis system. Available software solutions apply different strategies for consecutive image alignment and protein spot detection. While DeCyderTM performs spot detection on single gels prior to the alignment of spot maps, SameSpotsTM completes image alignment in advance of spot detection. In this study, the performances of DeCyderTM and SameSpotsTM were compared considering all protein spots detected in 2D DIGE resolved proteomes of three different environmental bacteria with minimal user interference. Proteome map-based analysis by SameSpotsTM allows for fast and reproducible abundance change determination, avoiding time-consuming, manual spot matching. The different raw spot volumes, determined by the two software solutions, did not affect calculated abundance changes. Due to a slight factorial difference, minor abundance changes were very similar, while larger differences in the case of major abundance changes did not impact biological interpretation in the studied cases. Overall, affordable fluorescent dyes in combination with fast CCD camera-based image acquisition and user-friendly image analysis still qualify 2D DIGE as a valuable tool for quantitative proteomics.

The introduction of two-dimensional gel electrophoresis (isoelectric focusing combined with SDS-PAGE; 2DE) around 40 years ago by O’Farrell [1975] allowed for the high-resolution separation of complex protein mixtures based on isoelectric point and molecular size. Although the development of immobilized pH gradients for first-dimension separation greatly enhanced the reproducibility of 2DE, gel-to-gel variation still required the use of technical replicates to encompass experimental variation [Görg et al., 2009]. Even though diverse postelectrophoretic fluorescence staining techniques (e.g., SYPRO® ruby or flamingo) already allowed 2DE with high sensitivity and a broad dynamic range, the challenge of gel-to-gel variation remained. In this respect, the concept of two-dimensional difference gel electrophoresis (2D DIGE), introduced in 1997, represented a major advance for quantitative 2DE [Ünlü et al., 1997]. 2D DIGE is based on the application of three mass- and charge-matched fluorescent cyanine-dyes for pre-electrophoretic protein labelling allowing for the separation of three different protein samples within a single gel. Thereby, gel-to-gel variation is avoided since exactly the same electrophoretic parameters are employed to the three samples. Furthermore, the use of an internal standard not only facilitated normalization of biological replicate samples, but also allowed for multiplexing due to the possibility of matching spots across multiple gels and, hence, samples [Gade et al., 2003; Beckett, 2012]. Since its introduction, 2D DIGE has been applied to all kinds of organisms and tissues [for overview see Cramer and Westermeier, 2012].

Following electrophoretic separation, gel images are visualized and digitalized originally using a 3-laser fluorescence scanner [Gade et al., 2003; Vormbrock et al., 2012] or, as recently introduced, specific CCD camera systems [e.g. Strijkstra et al., 2016]. The digital gel im ages are then processed by special software to determine quantitative differences in protein abundance. Several software solutions are commercially available, with De CyderTM (GE Healthcare), Delta2DTM (Decodon), and SameSpotsTM (TotalLab) being widely used. Although determining a similar parameter (i.e., abundance change), the way of raw-image processing, spot detection, and normalization differs between these software solutions. A “traditional” approach is applied by DeCyderTM using a co-detection algorithm for spot detection that combines information of all three, co-separated gel images to define a spot. Spot detection is performed for each gel individually, prior to the quantification and normalization of spot volumes and final spot matching to a master gel across the entire gel-set of an experiment (Fig. 1). In contrast, Delta2DTM and SameSpotsTM correct for electrophoretic migration differences in spot positions between gels by image warping, i.e., positional correction of gel images combining global and local image transforms prior to spot detection, generating a single fusion image (proteome map). This proteome map contains the information for all spots detected (no empty values) in the experimental gel set and is the basis for generation of a consensus spot pattern valid for all gels [Berth et al., 2007]. This workflow avoids the (time-consuming) necessity of reviewing the correctness of spot matching as in the case of DeCyderTM. Similar to DeCyderTM, normalization is performed prior to abundance profile analysis. A detailed comparison of the DeCyderTM and SameSpotsTM workflow is provided in Figure 1.

Fig. 1.

Comparative workflow of discrete steps during 2D DIGE analyses by SameSpotsTM and DeCyderTM. Overview of consecutive steps of the 2D DIGE analysis by SameSpotsTM (left) and DeCyderTM (right) with equivalent sections connected and color coded according to function: yellow, image upload and project creation; green, spot detection and matching; blue, spot exclusion during detection; purple, spot review; red, statistical analysis and report. For DeCyderTM, the different software modules are distinguished. PTM, posttranslational modification.

Fig. 1.

Comparative workflow of discrete steps during 2D DIGE analyses by SameSpotsTM and DeCyderTM. Overview of consecutive steps of the 2D DIGE analysis by SameSpotsTM (left) and DeCyderTM (right) with equivalent sections connected and color coded according to function: yellow, image upload and project creation; green, spot detection and matching; blue, spot exclusion during detection; purple, spot review; red, statistical analysis and report. For DeCyderTM, the different software modules are distinguished. PTM, posttranslational modification.

Close modal

So far, only a few studies have examined the influence of these different software approaches on the obtained quantitative data [Karp et al., 2008; Kang et al., 2009; Silva et al., 2010], albeit mostly focusing on a small subgroup of proteins (≤50) for detailed analysis. In this study, the influence of two state-of-the-art 2D DIGE image analysis software solutions, DeCyderTM and SameSpotsTM, on protein abundance profiles of three bacterial proteomes was investigated. For this purpose, all detected protein spots (730–4,183) of three environmental, proteomically well-studied model bacteria, i.e., Desulfobacula toluolica Tol2, “Aromatoleum aromaticum” EbN1, and Phaeobacter inhibens DSM 17395, were considered.

Dataset

To assess the influence of 2D DIGE image analysis on determined abundance differences, protein profiles of three proteogenomically well-studied environmental bacteria were analyzed: marine, sulfate-reducing deltaproteobacterium D. toluolica Tol2 [Wöhlbrand et al., 2013], terrestrial, denitrifying betaproteobacterium “A. aromaticum” EbN1 [Rabus et al., 2014], and marine, aerobic alphaproteobacterium P. inhibens DSM 17395 [Drüppel et al., 2014]. The gel set of each organism contained four gels in which protein extracts of four biological replicate samples were separated to account for biological variation [Zech et al., 2011]. While the test state varied for each bacterium (D. toluolica Tol2: toluene; “A. aromaticum” EbN1: 4-ethylphenol; P. inhibens DSM 17395: casamino acids), cells grown with fumarate served as the reference state in all cases. The internal standard was composed of equal amounts of all test and reference state samples per bacterium [Strijkstra et al., 2016]. This experimental setup yielded on average a total of 3,786–4,183 and 730–915 spots detected by DeCyderTM and SameSpotsTM, respectively, all of which were considered for the subsequent analyses (Table 1).

Table 1.

Comparison of the number of overall detected spots as well as shared or uniquely detected ones by DeCyderTM and Same­SpotsTM for the three studied bacteria (spot numbers are average values of 4 biological replicates)

 Comparison of the number of overall detected spots as well as shared or uniquely detected ones by DeCyderTM and Same­SpotsTM for the three studied bacteria (spot numbers are average values of 4 biological replicates)
 Comparison of the number of overall detected spots as well as shared or uniquely detected ones by DeCyderTM and Same­SpotsTM for the three studied bacteria (spot numbers are average values of 4 biological replicates)

Spot Detection and Master Gel Assignment

Prior to spot detection, the DeCyderTM software requires the user to specify an estimated number of spots. The DeCyderTM manual recommends setting the number of estimated spots to 2,500–10,000, for which reason four different settings were tested: (i) 1,500, (ii) 3,000, (iii) 5,000, and (iv) 10,000. Estimation of 1,500 spots yielded on average approximately 1,200 detected spots per gel, independent of the sample (online supplementary Fig. S1; see www.karger.com/doi/10.1159/000494083 for all online suppl. material). With 3,000 estimated spots, the number of detected spots increased to approximately 2,600, and with estimation of 5,000 or 10,000 spots the maximum of approximately 4,000 spots per gel and sample were detected. Notably, comparing the spot coordinates of the spot map estimating 1,500 spots with those estimating more, 86% of the spots did not change in position (one-pixel tolerance) and revealed a difference in the raw spot volume of only 8% (data not shown). The number of detected spots directly influences the DeCyderTM data analysis, since the gel with the most spots detected is selected as the master gel that is subsequently used for spot matching of all other images of a gel set. The assigned master gel remained constant with ≥3,000 estimated spots, except for the P. inhibens DSM 17395 sample (≥5,000). Since no user influence on spot detection is possible in SameSpotsTM, and user-defined parameters were reported to influence the degree of statistical confidence of DeCyderTM analysis [Gade et al., 2003], an estimated spot number of 5,000 was used for DeCyderTM spot detection to minimize user bias.

Applying these settings, on average 4,000 spots per sample and gel were detected by the DeCyderTM software, while SameSpotsTM detected only 730–915 (Table 1; Fig. 2). Notably, only < 50% (on average 1,845 spots) of the DeCyderTM-detected spots were present in all gels of the respective gel set (online suppl. Fig. S1), which was twice as many as compared to SameSpotsTM, where all spots are present in all gels due to the warping routine of the algorithm. Furthermore, 25% of the SameSpotsTM-assigned spots (in the case of P. inhibens DSM 17395 even 30%) were interpreted by DeCyderTM as two or more spots (up to 9) when comparing spot boundaries (online suppl. Fig. S2), indicating that these numbers are falsely high. This assumption is supported by detection of only 25% of the DeCyderTM spots in all gels of an Erwinia carotovora proteome analysis applying the SameSpotsTM software [Karp et al., 2008]. Notably, in the present study, 73–80% of all SameSpotsTM spots were also assigned by the DeCyderTM software (including all “clearly visible” spots; Fig. 2), indicating a similar spot data set for analysis. Previously reported merging of apparently different spots to one spot (i.e., omitted splitting of overlapping spots) in the case of SameSpotsTM [Karp et al., 2008; Morris et al., 2010] could not be confirmed in this study, most likely due to the more advanced version of the Same­SpotsTM software (2006 vs. 2017).

Fig. 2.

Total number of detected spots and their corresponding abundance characteristics assigned by SameSpotsTM and DeCy­derTM for D. toluolica Tol2, “A. aromaticum” EbN1, and P. inhibens DSM 17395. For each bacterium and software, the number of detected spots is given in gray, increased abundance in green, decreased abundance in red, and unchanged abundance in orange (significantly changed abundance: ANOVA p value ≤1 × 10–4 and fold change ≥1.5 or ≤–1.5). The number of spots of the reduced dataset similarly assigned by both software solutions is indicated in lighter shading. The number of spots present in both software solutions, but not fulfilling the significance criteria (i.e., ANOVA p value or abundance change) in the respective software equivalent is indicated by hatching (white).

Fig. 2.

Total number of detected spots and their corresponding abundance characteristics assigned by SameSpotsTM and DeCy­derTM for D. toluolica Tol2, “A. aromaticum” EbN1, and P. inhibens DSM 17395. For each bacterium and software, the number of detected spots is given in gray, increased abundance in green, decreased abundance in red, and unchanged abundance in orange (significantly changed abundance: ANOVA p value ≤1 × 10–4 and fold change ≥1.5 or ≤–1.5). The number of spots of the reduced dataset similarly assigned by both software solutions is indicated in lighter shading. The number of spots present in both software solutions, but not fulfilling the significance criteria (i.e., ANOVA p value or abundance change) in the respective software equivalent is indicated by hatching (white).

Close modal

Differences in Protein Abundance Profiles

Similar criteria were applied to classify a spot to be significantly changed in abundance for both software solutions (i.e., a fold-change ≥∣1.5∣ and an ANOVA p value < 0.05). To allow for direct comparison of different parameters of corresponding DeCyderTM and SameSpotsTM spots, respectively, each DeCyderTM spot was manually matched to its corresponding SameSpotsTM spot equivalent. Here, the spot intensity maximum (indicated by DeCyderTM) was used as the decisive parameter and the spot with the largest spot volume (internal standard) was chosen in the case of multiple DeCyderTM spots per SameSpotsTM spot (online suppl. Fig. S2A).

For all three studied bacterial species, more spots were assigned to be significantly increased in abundance by DeCyderTM as compared to SameSpotsTM (121–207 vs. 91–108, respectively; Fig. 2). Out of these, 38–82 were similarly assigned by both software solutions. It should be noted that assignment of multiple spots by DeCyderTM within the spot boundaries of a single SameSpotsTM spot (online suppl. Fig. S2A) may artificially increase the respective number of DeCyderTM spots, given they fulfill the respective criteria (i.e., significantly increased abundance in this case). However, for 17–25% of the DeCyderTM spots with increased abundance, no significant abundance change or a decreased abundance was assigned by SameSpotsTM for the respective spot. Contrariwise, 14–45% of the SameSpotsTM spots with increased abundance were not similarly assigned by DeCyderTM. In the case of spots with decreased abundance, the number was lower for DeCyderTM than for SameSpotsTM (27–74 vs. 50–84), except for D. toluolica Tol2 (121 vs. 86), which had similar assignment for 9–53 spots (Fig. 2). Out of these spots with decreased abundance, 5–25% (DeCyderTM) and 21–64% (SameSpotsTM), respectively, were differently assigned. Overall, a rather large number of spots (353–523) was similarly assigned to be not significantly changed in abundance, accounting for 10–13% (DeCyderTM) and 64–68% (SameSpotsTM) of the nonregulated spots, respectively. The comparably low share in the case of DeCyderTM is mainly due to the > 4.5-fold higher number of total detected spots, most of which do not have a SameSpotsTM equivalent, as well as spot fractionation, i.e., assignment of multiple DeCyderTM spots per SameSpotsTM spot. In general, spots that were assigned to be significantly changed in abundance by only one of the two tested software types either just did not meet the abundance change or the ANOVA threshold (seldom both) in the case of the other image analysis system.

Software Impact on Abundance Changes

Differences in determined abundance changes were analyzed using a reduced dataset containing only spots present in both gel sets, i.e., 719 for D. toluolica Tol2, 666 for “A. aromaticum” EbN1, and 565 for P. inhibens DSM 17395, covering 73–80% of all SameSpotsTM-detected spots. For each protein spot, its SameSpotsTM abundance change was plotted versus the DeCyderTM equivalent (Fig. 3a, c, e). Linear regression analysis revealed that the SameSpotsTM values were 0.55- to 0.72-fold lower as compared to DeCyderTM (coefficient of determination 0.69–0.92). However, considering only spots meeting the 95% confidence interval (assuming a 1: 1 relation), the deviation was reduced to 0.69- to 0.90-fold (coefficient of determination 0.71–0.93), although only 20–31 spots were not considered (out of 565–719). Hence, SameSpotsTM apparently calculates an 8–30% smaller change in abundance as compared to DeCyderTM and thus larger fold change differences in the case of a large abundance changes. This finding contrasts with the reported large difference in DeCyderTM- and SameSpotsTM-calculated abundance changes of three selected spots [Kang et al., 2009], but is in line with determined subtle differences of 50 spots in the case of the E. carotovora proteome [Karp et al., 2008].

Fig. 3.

Inter-software correlation of spot attributes: abundance change and raw volume dot plot. Correlation analysis was performed with the reduced data set (i.e., one DeCyderTM spot is assigned to the equivalent SameSpotsTM spot) for the change in abundance (a, c, e) and the raw volume (b, d, f) for D. toluolica Tol2, “A. aromaticum” EbN1, and P. inhibens DSM 17395. For the abundance change analysis (left column), the 95% interval (assuming a 1: 1 relation) is indicated in light green. Spots not meeting this 95% criterion are highlighted in red. The coefficients of determination and the line of best fit are given for the total dataset (including outliers, red) and those meeting the 95% criterion (blue). For the raw volume plots (right column), spots not meeting the 95% criterion in the abundance change analysis are also highlighted in red.

Fig. 3.

Inter-software correlation of spot attributes: abundance change and raw volume dot plot. Correlation analysis was performed with the reduced data set (i.e., one DeCyderTM spot is assigned to the equivalent SameSpotsTM spot) for the change in abundance (a, c, e) and the raw volume (b, d, f) for D. toluolica Tol2, “A. aromaticum” EbN1, and P. inhibens DSM 17395. For the abundance change analysis (left column), the 95% interval (assuming a 1: 1 relation) is indicated in light green. Spots not meeting this 95% criterion are highlighted in red. The coefficients of determination and the line of best fit are given for the total dataset (including outliers, red) and those meeting the 95% criterion (blue). For the raw volume plots (right column), spots not meeting the 95% criterion in the abundance change analysis are also highlighted in red.

Close modal

Spot fractionation by DeCyderTM, i.e., assignment of 2 or more spots per SameSpotsTM spot, may indicate generally larger spot volumes by the latter. However, most spots of the reduced dataset revealed a larger spot volume assigned by DeCyderTM (61–83%; Fig. 3b, d, f), although SameSpotsTM assigned a pronouncedly larger volume to some spots (most prominent in the case of D. toluolica Tol2). Notably, spots with large differences in calculated abundance changes revealed only average differences with respect to spot volume (red dots in Fig. 3).

Both software solutions apply normalized spot volumes for the calculation of abundance changes. While the procedure is not described in detail for DeCyderTM, the normalization procedure of SameSpotsTM is replicable. To assess the influence of normalization on abundance change calculation, a custom algorithm was constructed reflecting the SameSpotsTM procedure. Comparing the custom calculated normalization factor with that of the SameSpotsTM software revealed an average difference of ≤0.01. Hence, the custom calculation may be used to represent the SameSpotsTM normalization procedure. By applying the raw spot volumes determined by DeCyderTM in this calculation, the obtained abundance changes were similar to those calculated with SameSpotsTM raw volumes (median deviation: D. toluolica Tol2: 0.01; “A. aromaticum” EbN1: 0.02; P. inhibens DSM 17395: 0.08), demonstrating that the differences in software-determined spot volumes do not affect the calculated abundance change.

Software Influence on Biological Interpretation

In general, due to the apparently factorial differences in calculated abundance changes, both software solutions determine highly similar fold changes in the case of proteins with small changes in abundance. For example, a similar increase in abundance was determined for two proteins (out of seven identified) involved in anaerobic degradation of toluene by D. toluolica Tol2, i.e., 3.6-fold and 1.9-fold in the case of 2-(hydroxy[phenyl]methyl)-succinyl-CoA dehydrogenase (BbsC) and phenylitaconyl-CoA hydratase (BbsH), respectively (online suppl. Fig. S3). However, minor deviations in the case of proteins with larger fold changes (e.g., 27.7-fold vs. 22.2-fold in the case of benzylsuccinate synthase subunit A [BssA] for DeCyderTM and SameSpotsTM, respectively), should not be paid too much attention, since the determined fold change is still pronounced, and the relative change in abundance is the same for other test states of the experiments. Hence, an overall similar set of significantly changed proteins will be detected by both software solutions, allowing for a similar biological interpretation, as exemplified for the toluene degradation proteins of D. toluolica Tol2 (online suppl. Fig. S3) and previously reported for a smaller set of proteins [Karp et al., 2008]. It is noteworthy that the impact of sample generation, handling, and preparation, as well as the electrophoresis and image acquisition (i.e., digitalization system), is presumably more pronounced than that of the analysis software [Silva et al., 2010; Zech et al., 2011; Strijkstra et al., 2016].

Although the image upload and project generation is comparable for DeCyderTM and SameSpotsTM, handling of the latter is more intuitive for the user due to fewer instances of user intervention (with respect to possibility and necessity) as well as its nonmodular structure allowing for a clear and coherent workflow, as reported previously [Silva et al., 2010]. Furthermore, markedly less time has to be invested for quantitative analysis since the warping technique combined with image alignment prior to coherent spot detection supersedes the time-consuming spot matching recommended in the case of DeCyderTM. However, the procedure of manual spot boundary changes and addition of not automatically detected spots is easier in the case of DeCyderTM, with respect to spot review (2D and 3D) and manipulation. The discovered small, factorial difference in calculated abundance changes is most likely evoked by the DeCyderTM normalization procedure, possibly involving other, unknown implemented processes.

Overall, even 40 years after the invention of 2DE and 20 years since 2D DIGE, the latter represents an affordable and valuable tool for high-resolution, quantitative differential proteomics. This is because of: (i) reduced costs of fluorescent dyes (due to meanwhile multiple vendors), (ii) time-saving and easy to use advanced CCD camera systems, as well as (iii) intuitive workflow and fast image analysis by warped, proteome map-based software, allowing for (iv) determination of highly similar, reproducible, and user-independent protein abundance changes.

2D DIGE and Image Acquisition

Inter-software comparison was performed using three distinct 2D DIGE gel sets of three different environmental bacteria: D. toluolica Tol2, “A. aromaticum” EbN1, and P. inhibens DSM 17395. 2D DIGE was performed using 24-cm IPG-strips and image acquisition was performed by a charge-coupled camera system as described in detail by Strijkstra et al. [2016].

General Software Comparison

In this study, image analysis was performed using DeCyderTM version 7.2 (GE Healthcare, Munich, Germany) and SameSpotsTM version 5.0.1.0 (TotalLab, Newcastle upon Tyne, UK). Comparison of the different steps from image analysis to spot abundance data calculation (Fig. 1) was based on the information provided in the user manual of DeCyderTM (version 7.0) and the SameSpotsTM DIGE tutorial, company homepage, and personal communication (May 2017).

Image Processing

The strain-specific gel sets were imported into the respective software and processed with settings set to defaults or automatic procession. User interference was kept to a minimum. However, it was necessary for the user to provide an estimated number of spots with the DeCyderTM software. For each gel set, the estimated number of spots was set to 5,000 in the Batch processor module. Spot normalization and quantitation was not altered manually in either of the software types. The following information was collected from both software solutions: gel images with spot boundaries, software inherent spot number, spot coordinates, normalization factor, spots excluded from the normalization process, raw and normalized spot volume, fold change, and one-way ANOVA p values.

Inter-Software Spot Map Matching

For each gel set, the internal standard gel image of the DeCyderTM master gel (automatically determined) with indicated spot boundaries was employed to create an overlay with the analogous spot map of SameSpotsTM (online suppl. Fig. S2). Based on this overlay image, spots present in both spot maps were determined and matched, applying the “center of mass” denoted by DeCyderTM as the decisive parameter. The latter is defined by the DeCyderTM software as the optimal picking location for a respective protein spot.

The DeCyderTM software does not request the manual checking of the spot matching across all spot maps of a gel set. However, it is recommended for spots with assigned decreased or increased abundance. Hence, a second list of DeCyderTM protein spots with corresponding SameSpotsTM spot equivalents was created after manual spot confirmation in DeCyderTM. Three types of DeCyderTM spots were chosen for manual confirmation: (i) spots with a SameSpotsTM equivalent, (ii) spots with increased or decreased abundance (in either software), and (iii) spots with their “center of mass” close to, but not inside the respective SameSpotsTM spot boundary. In contrast to DeCyderTM, the SameSpotsTM software does not request any spot confirmation or correction. Therefore, the same SameSpotsTM data set was used for both inter-software spot matching lists.

In several cases, more than one DeCyderTM spot could be assigned to a distinct SameSpotsTM spot. To allow for direct comparison of distinct spot attributes per spot, the DeCyderTM spot with the highest raw spot volume of the internal standard (mean of the gel set) was assigned.

Comparison of Spot Statistics

The two different software packages were compared based on the raw spot volume as well as on computed values, i.e., normalized spot volume, fold change (SameSpotsTM), or the average ratio (DeCyderTM), respectively, and one-way ANOVA p values. While decreased abundance is indicated by the DeCyderTM average ratio as a negative value, the SameSpotsTM fold change is generally positive. To allow for direct comparison of decreased abundances, respective SameSpotsTM spots were targeted in hierarchical grouping of the dendrogram in the software and converted to the corresponding negative value. Spots significantly changed in abundance had a fold change ≥1.5 or ≤–1.5 and an ANOVA p value ≤1 × 10–4.

Effect of Normalization

The two software packages apply different algorithms to normalize the raw volume data of the test and reference state, respectively, to the internal standard. SameSpotsTM uses an iterated median and median absolute deviation approach to calculate upper and lower limits for the log volume ratio.

To model the calculation of the limits by the SameSpotsTM software, the median and its deviation can be calculated from the log volume ratio of a gel pair (test state and internal standard or reference state and internal standard, respectively). Subsequently, this step is iterated on the log volume ratio of all spots below the negative median deviation, yielding a new positive and negative median deviation. The latter can then be used as estimation of the lower limit calculation by the SameSpotsTM software. The upper limit can be modelled in the same way; however, here the positive median deviation of the spots initially above the positive median deviation is used for the calculation of the normalization factor.

Ten to the power of the mean of the log volume ratio of all spots inside these limits is then used as the normalization factor. The normalized volume may then be calculated by dividing the raw volume of the test state and reference state, respectively, by the raw volume of the internal standard and subsequent multiplication with the normalization factor. Finally, the abundance change is determined by division of the normalized volume of the test state by the reference state. In a case where the normalized volume is larger in the reference state, the latter is divided by the test state.

This normalization procedure was reconstructed in the R project software and RStudio (version 1.1.383). Unfortunately, a detailed description of the normalization procedure of DeCyderTM is not publicly available.

The authors are grateful to M. Prochnow (Oldenburg) for support during R programming and C. Feenders (Oldenburg) for interesting discussions. This study was supported by the Deutsche Forschungsgemeinschaft (SFB TRR 51) and the Niedersächsisches Ministerium für Wissenschaft und Kultur (Promotionsprogramm “The Ecology of Molecules – EcoMol”).

The authors have no conflicts of interest to declare.

1.
Beckett P: The basics of 2D DIGE. Methods Mol Biol 2012; 854: 9–19.
2.
Berth M, Moser FM, Kolbe M, Bernhardt J: The state of the art in the analysis of two-dimensional gel electrophoresis images. Appl Microbiol Biotechnol 2007; 76: 1223–1243.
3.
Cramer R, Westermeier R: Difference Gel Electrophoresis (DIGE). New York, Springer, 2012.
4.
Drüppel K, Hensler M, Trautwein K, Kossmehl S, Wöhlbrand L, Schmidt-Hohagen K, Ulbrich M, Bergen N, Meier-Kolthoff JP, Göker M, Klenk HP, Schomburg D, Rabus R: Pathways and substrate-specific regulation of amino acid degradation in Phaeobacter inhibens DSM 17395 (archetype of the marine Roseobacter clade). Environ Microbiol 2014; 16: 218–238.
5.
Gade D, Thiermann J, Markowsky D, Rabus R: Evaluation of two-dimensional difference gel electrophoresis for protein profiling. Soluble proteins of the marine bacterium Pirellula sp. strain 1. J Mol Microbiol Biotechnol 2003; 5: 240–251.
6.
Görg A, Drews O, Luck C, Weiland F, Weiss W: 2-DE with IPGs. Electrophoresis 2009; 30:S122–S132.
7.
Kang Y, Techanukul T, Mantalaris A, Nagy JM: Comparison of three commercially available DIGE analysis software packages: minimal user intervention in gel-based proteomics. J Proteome Res 2009; 8: 1077–1084.
8.
Karp NA, Feret R, Rubtsov DV, Lilley KS: Comparison of DIGE and post-stained gel electrophoresis with both traditional and SameSpots analysis for quantitative proteomics. Proteomics 2008; 8: 948–960.
9.
Morris JS, Clark BN, Wei W, Gutstein HB: Evaluating the performance of new approaches to spot quantification and differential expression in 2-dimensional gel electrophoresis studies. J Proteome Res 2010; 9: 595–604.
10.
O’Farrell PH: High resolution two-dimensional electrophoresis of proteins. J Biol Chem 1975; 250: 4007–4021.
11.
Rabus R, Trautwein K, Wöhlbrand L: Towards habitat-oriented systems biology of “Aromatoleum aromaticum” EbN1: chemical sensing, catabolic network modulation and growth control in anaerobic aromatic compound degradation. Appl Microbiol Biotechnol 2014; 98: 3371–3388.
12.
Silva E, O’Gorman M, Becker S, Auer G, Eklund A, Grunewald J, Wheelock AM: In the eye of the beholder: does the master see the SameSpots as the novice? J Proteome Res 2010; 9: 1522–1532.
13.
Strijkstra A, Trautwein K, Roesler S, Feenders C, Danzer D, Riemenschneider U, Blasius B, Rabus R: High performance CCD camera system for digitalisation of 2D DIGE gels. Proteomics 2016; 16: 1975–1979.
14.
Ünlü M, Morgan ME, Minden JS: Difference gel electrophoresis: a single gel method for detecting changes in protein extracts. Electrophoresis 1997; 18: 2071–2077.
15.
Vormbrock I, Hartwig S, Lehr S: Multifluorescence 2D gel imaging and image analysis. Methods Mol Biol 2012; 854: 21–30.
16.
Wöhlbrand L, Jacob JH, Kube M, Mussmann M, Jarling R, Beck A, Amann R, Wilkes H, Reinhardt R, Rabus R: Complete genome, catabolic sub-proteomes and key-metabolites of Desulfobacula toluolica Tol2, a marine, aromatic compound-degrading, sulfate-reducing bacterium. Environ Microbiol 2013; 15: 1334–1355.
17.
Zech H, Echtermeyer C, Wöhlbrand L, Blasius B, Rabus R: Biological versus technical variability in 2-D DIGE experiments with environmental bacteria. Proteomics 2011; 11: 3380–3389.