Western blotting represents a powerful technique for the semi-quantitative determination of protein expression and the detection of protein modifications like phosphorylation and ubiquitylation. It is still one of the most commonly used methods in modern cell biology laboratories due to its high specificity and sensitivity.
As imaging software like Photoshop became so widely available, it is now quite simple to adjust or modify digital image files – sometimes in a way that constitutes inappropriate changes to the original data (see Figure 1 below).

Figure 1: The colour-coded labelling of bands indicates duplication and reuse of different gel sections. This doctored image appeared in a scientific publication, which was later corrected as the data from original gels were incorrectly compiled or modified by the first author (modified from https://pubpeer.com/publications/20504287).

However, image manipulation based on fabrication and falsification most likely accounts for only a minority of cases and it is estimated that less than 0.1% of all publications need to be retracted due to scientific fraud and misconduct (including all different techniques, not only Western blotting).
A much bigger concern, affecting many more scientists, scientific studies, time and resources is the poor reproducibility of western blotting-based experiments due to lack of antibody validation, missing information and controls, technically sloppy execution and ‘beautification’.
 
Cropping:
On the one hand, cropping gels can improve ‘the clarity and conciseness of the presentation’ but it has to be ensured that no information gets lost. Quite often, a blot is cut into a narrow strip or cropped around individual bands (see Figure 2, panel on the right) to remove ugly (beautification), unexplainable or confusing areas.

Figure 2: Comparison of the breast cancer marker protein Ezh2 expression levels between different cell lines. The full western blot (ranging from 30-200 kDa) is shown on the left, including specific and unspecific proteins detected by the antibody. The right panel represents the area typically selected for publication (modified from https://media.cellsignal.com/pdf/4905.pdf).

Given the example shown in Figure 2, to follow Good Research Practice recommendations, the full blot on the left including molecular weight markers should be shown and the two key questions should be addressed: How was the antibody validated and the specificity controlled? What is the origin of the band below 80 kDa as the band intensity is inversely proportional to the band around 100 kDa?
Unfortunately, it is common practice to reduce the full blot (Figure 2, left) to a small section (Figure 2, right) without molecular weight markers and to establish antibody specificity by just labelling the ‘correct’ band.
Not only can this procedure mislead the reader when judging the results but it can also hide essential information pointing to real biological insight. Importantly, it may also lead to a false impression of antibody specificity, and hence the level of controls required to establish band identity. Therefore, uncropped, full images of Western blots should be submitted as supplementary material wherever possible.
 
Splicing:
It is against Good Research Practice to mix and match bands from various gels that were run at different times or exposed for different amounts of time and generate data that appear to have been run simultaneously and under identical conditions. For readers to know whether they can compare different bands on the same blot, they need to know whether they came from the same gel.
If vertical lines are visible between lanes (Figure 3), this normally represents some unnatural appearances and image manipulation and should be treated with caution.

Figure 3: Vertical lines between lanes 2/3, 4/5 and 6/7 indicate that sections from different gels were spliced together (modified from http://onlinelibrary.wiley.com/wol1/doi/10.1111/j.1365-2559.2011.04146.x/full).

If splicing different gels together is unavoidable, a black line or gap should be left so that it is clear that the parts had separate origins.
 
It is critical to follow Good Research Practices when performing and, equally important, presenting data generated by SDS-PAGE and Western blotting and for every adjustment that is done, it is always important to ask ‘Is the image that results from these adjustments still an accurate representation of the original data?’