Changing the nature of separation

Label-Free Intrinsic Imaging (LFII) is a new method of separation that promises substantial advantages over standard technologies currently in use. Here Stuart Hassard explains how deltaDOT got to grips with the technique 

LFII is said to deliver excellent data quality – including high resolving power, high reproducibility, quantification, signal-to-noise ratios and intrinsically digital data – and can do so without the use of any labels.

LFII utilises the optical properties of molecules, as defined by the Beer-Lambert law, in combination with multipixel data acquisition and complex signal processing algorithms. The LFII approach facilitates dramatic improvements over conventional UV detection technologies in terms of resolution, reproducibility and direct quantification resulting in significantly enhanced quality of output.

The absorption of light by a solution with a concentration, c, is described by the Beer-Lambert law,
                                              

where I is the intensity once the light has passed through the sample, I0 is the initial unobstructed intensity, p is the path length the light travels through the solution, and k(λ) is the wavelength-dependent absorption coefficient.

A concentration detection limit is the point at which analyte concentrations below this limit cannot be detected. A high concentration detection limit means lower transmission levels, which leads to a high detection limit of the instrument in question. A lower instrument detection limit is an indication of improved signal-to-noise ratios.

A significant limitation in current Capillary Electrophoresis (CE) systems is that when UV absorbance detection is used, only relatively poor data quality is attainable. deltaDOT is offering improved data quality by the application of LFII to CE. At the heart of deltaDOT’s LFII proprietary algorithms is the principle that accuracy can be improved by performing multiple measurements of an observable event. A principal method of increasing data quality is through the use of signal averaging. Signal averaging using conventional computation methods would consume a huge amount of processing power and therefore require significant computation resources.

Signal averaging is not practicable in conventional CE runs because the migration times are not sufficiently reproducible to perform repetitive runs in order to acquire more data. Furthermore, the time required to repeat enough runs to increase sufficiently the signal-to-noise ratio, and thus data quality, would be unworkable. LFII allows signal averaging to be achieved during a single run, by means of multiple point detection along the capillary. The analyte signal from each detector may then be averaged to increase the signal-to-noise ratio. The signal-to-noise enhancement will be equal to the square root of the number of diode detectors, if the detectors are shot noise- or white (random) noise-limited.

The principle of vertexing demonstrated on DNA ladder data. Each base reaction is identified by its injection point. The signal processing allows Virtual Colour to identify the base reaction and generate the sequence.

Separations are dynamic procedures in which progressive band broadening processes (temperature, diffusion) are present. Consequently, as resolution subsequently decreases throughout a CE run, a linear photodiode array (PDA) is used for the multiple detection points required. This reduces effects due to electrophoretic distortion effects, as the diodes are closely related. At deltaDOT, consecutive photodiode pixel detectors are employed, producing an array arranged linearly along the capillary at a post-separation imaging position. Each diode produces an independent electropherogram sampling at up to 60 Hz, thus generating thousands of electropherograms for signal averaging. These measurements are combined with the space-time injection origin of the sample as a constraint to produce the best possible noise reduction. At high light intensities, as will be the case for absorbance detection, the linear PDA is shot noise-limited. Shot noise is commonly defined as the collective noise produced by the random generation of photons from the excitation source and the random generation of electrons in the photodiode. The shot noise for a fully saturated diode is the square root of the saturation charge for each diode. In LFII the theoretical signal-to-noise improvement is the square root of 512, which is 22.6.

A Generalised Separation Transform (GST) algorithm is designed to retain maximum information on spectra shape as well as quantification. The GST algorithm maximises signal-to-noise in a ‘natural’ way, that is, with an approach that maximises signal-to-noise but retains shape information of the analyte peak. GST is less sophisticated than the EVA approach, discussed later, and could usefully be employed as an analysis front-end.

The Equiphase Vertexing Algorithm (EVA) analyses the electropherograms and reduces them to a set of space-time points corresponding to detected absorption peaks. The resulting set of signals clearly shows the sample bands moving through the system. They also show any irregularities in the gel, e.g. bubbles, as signals that move with anomalous velocities. This has proved to be a useful monitor of run quality. The signals can be associated with, and used to improve knowledge of the injection vertices. Combining the signals with their associated vertices, results in clearly defined and well-quantified bands. It can also be used to differentiate between different injections (Figure 2) allowing multiple injection to be performed in a single electrophoretic run.

Both GST and EVA algorithms are designed to maximise the resolution (the algorithmic or “virtual” resolution) and reproducibility of the data. The standard quantification is based on peak height, with peak area available as an alternative measure. Through peak detection, EVA allows for better resolution relative to GST. This is due to the fact that GST can smudge peaks if an analyte velocity does not appear constant through out its separation. This can be due to optics, diffusion, or other phenomena. In contrast, GST retains signal shape as there is no prior peak detection present in the algorithm. Since all raw and processed data are kept other analyses can be performed.

Multipoint detection and label-free photonic imaging have mathematical and physicochemical consequences that result in very important practical advantages. First, the geometric method of analysis makes it possible to use the powerful signal processing algorithms (GST and EVA) that are computationally highly efficient, so that they are able to run in real time on a standard PC. Secondly, the ability to bring powerful signal processing algorithms to bear allows the use of very low signal levels, which permits the use of photonic imaging rather than label detection. Thirdly, photonic imaging has the advantage over the use of labels that the strength of the signal depends on the length of the path of light through the sample. This is in contrast to the fact that the strength of a signal coming from a label is mathematically proportional to the effective volume of the label. This means that with decreasing sample size, the strength of the signal from a label falls as the inverse cube. Thus the strength of photonic signal falls far more slowly (linearly, rather than with the inverse cube) as sample amount is reduced, giving LFII a powerful advantage in detecting even minute amounts of sample. If a labeled system and an LFII system are both reduced in size by a factor 10, then the signal to noise of the labeled system drops 100 times more than that of the LFII.         

Separations of proteins, peptides and nucleic acids are routinely carried out using 1D-GE or HPCE. The use of 1D-GE is attractive because of its low cost and high throughput, but is a crude and unreliable method with many drawbacks, including slow speed and difficulty in quantifying separated components. Capillary electrophoresis instruments were introduced to the market to address many of the shortcomings of 1D-GE, but their performance characteristics have still left users requirements partially fulfilled. A newer generation of lab-on-a-chip instruments offer higher throughput but with only fair data quality and with significant consumables costs.

deltaDOT’s HPCE will compete directly with existing HPCE systems and provides superior sensitivity, signal to noise and reproducibility

deltaDOT’s LFII technology deltaDOT’s LFII technology applies signal detection, noise reduction and data handling algorithms originally developed to track subatomic particles in high energy physics, to molecular biology, giving improvements in analysis. The key elements of the system are: 1. accurately measuring the velocity of the analyte as it moves through the matrix, rather than its final position at the end of the experiment. 2. detecting the movement of analytes directly using molecule specific ultraviolet wavelength energy, rather than any sort of label or stain. 3. applying powerful signal-processing algorithms to track and identify the individual separating components. LFII is applied to deltaDOT’s Raptor range of analytical instruments including the Peregrine benchtop HPCE analyser (Figure 1), providing label free separations of proteins, glycoproteins, peptides, nucleic acids, bacteria, virus and small molecules. Also the Merlin high resolution DNA analysis system, for short run sequence on demand for QA/QC and resequencing applications and Short Tandem Repeat analysis in Forensics.

By Stuart Hassard. Stuart worked at the Department of Medicine at the University of Cambridge. When, with his elder brother Dr John Hassard of Imperial College London, he conceived the core technology that would lead to the creation of the spin-out company deltaDOT Ltd. He is currently head biologist in deltaDOT.