A brief guide to quantitative analysis of electrophoresis gels

1st April 2013

Dr Paul N Goulding discusses some of the important aspects of getting anumbers from pictures' ­ that is quantitative gel analysis.

Electrophoresis gels of DNA or protein separations are often treated with fluorescent staining materials for visualisation of the bands. It is also common to acquire digitised images using charge coupled device (CCD) camera-based gel documentation equipment, such as UVIdoc or UVIpro. These images can then be analysed using computer software in order to extract quantitative data such as band volumes and (therefore) band quantities (Fig. 1).

The type of analysis possible from araw' quantitative band data depends on the relationship between the amount of sample constituting a band and the fluorescence emitted by the stained band. This in turn, depends on the binding behaviour of the stain used. For DNA, it is generally assumed that intercalating dyes such as ethidium bromide are indiscriminate (random) in their binding behaviour, showing no preference for size or structure of the DNA molecules. Therefore, DNA quantity standards can be used to calculate quantities of material in other bands by calculating the relative volumes of the standard and the unknown. For proteins, the more diverse primary and secondary structure of different molecules can lead to variable binding of the fluorescent dye or label. This must be taken into account when performing quantification experiments and may mean that only bands of the same (or very similar) protein species are valid for use as reference standards in quantity calculations.

Quantification of electrophoresis bands is possible owing to the linear response of CCDs to the fluorescent output of the bands. A digitised gel image therefore comprises an array of picture elements (pixels) with numerical intensity values accurately representing the amount of light emitted by the corresponding points on the gel. A digitised band is most accurately quantified by adding the intensities of all constituent pixels to give a three dimensional band avolume'. Less accurate quantification methods such as measuring the area of a section through the band are sometimes used. However, these do not take account of irregular and asymmetrical band shapes which can cause significant errors (Fig. 2) and should therefore be avoided. The actual process of band quantification by software, such as UVIband, usually has the overall scheme shown in Fig. 3.

The first three stages, leading to generation of the raw volume data, are usually carried out automatically by the software. This is beneficial in that it saves work on the part of the user while minimising user dependence of the calculated data by standardising the methodology and parameters used in the process. However, despite the obvious benefits of automation, it is important to realise that software is not infallible. Indeed the objective nature of software, though enhancing standardisation of the analysis process, can occasionally lead to errors, particularly in band detection and definition. In other words, the ability of software to distinguish between eg, a band and a smear or artefact which hasband-like appearance is limited by the band recognition parameters programmed into it.Since these can never be perfect, the judgement and experience of the user is vital in these cases.

Finally, a brief discussion of background subtraction, potentially the most difficult process of the analysis scheme. There are many methods, a number of them suffering from severe drawbacks such as: much user input required; introduces user dependence into the results; performed one lane at a time: very time consuming; sensitive to small baseline variation: a aspike' or ablip' in the baseline, statistically insignificant, may cause a significant variation in the adjusted baseline.

Clearly a method is required which combines a high level of automation with a reliable means of baseline definition, while treating different parts of the gel independently. Such a method is the arolling ball' process in which the position of the baseline is determined by the position of a ball which rolls underneath the 3D surface of the image. The amount of baseline subtraction in this process is influenced by the size of the ball chosen, which can be defined by the software (but edited if necessary by the user). Reliable, systematic background subtraction is vital since very few gels have background levels which are consistent throughout the gel and it is common for bands to overlap, making determination of the baseline position far from trivial.

The overriding conclusion is that automated analysis software has many advantages, but it has limited powers of judgement. It is this vital element along with experience, which the user can bring to an analysis, often making the difference between good and bad data.


Dr Paul N Goulding is with UVItec Ltd, Cambridge, UK.





To receive the Scientist Live weekly email NewsBrief please enter your details below

Twitter Icon © Setform Limited