UltraPlex™ FingerPrint Technology
Molecular fingerprinting is a new approach to medical diagnostics. Fingerprints based on genes or proteins, and generated by expensive equipment such as NMR machines or mass spectrometers, have shown tremendous promise in research laboratories. To date no practical and economic method has previously been devised for generating such profiles. However, UltraPlex™ now makes delivery of revolutionary profiling diagnostics, in a convenient format and at an acceptable cost to users, a reality.
Pronostics uses its proprietary UltraPlex™ platform technology, antigen libraries, and pattern recognition software to create biological fingerprints focusing on profiling the complete immune system of an individual from a single, simple, blood sample. By measuring the levels of thousands of chemicals in a blood sample simultaneously it is possible to construct a profile of an individual which can be used to safely and accurately diagnose a wide range of different diseases.
The Principle of FingerPrint Technology
The principle of FingerPrint technology is very simple: if you measure enough things about an individual you can generate a profile (or molecular fingerprint) which is sufficiently detailed to allow you to diagnose virtually any disease from careful examination of that profile.
Pronostics are not the first people to recognise that this approach (called “multivariate diagnostics”) offers the potential to revolutionise in vitro medical diagnostics by providing high sensitivity and specificity tests based on a simple blood sample for many prevalent diseases where it is currently impossible to do so. Indeed, most of the opinion leaders in clinical diagnostics recognise the vast potential of this approach. But Pronostics scientists were the first people to make such a profiling diagnostic test work in practice, and Pronostics remains the only company with a genuinely low cost method of generating such a profile.
Working in collaboration with Prof. Jeremy Nicholson and his team at Imperial College, University of London, UK, our scientists, Dr David Grainger and Dr David Mosedale, obtained a molecular profile using NMR spectroscopy; a blood sample is placed inside an NMR spectrometer, which generates a metabolic profile containing information about hundreds of metabolites (such as sugars and fats) present in the sample. We chose NMR spectroscopy for our proof-of-principle studies because, although it is expensive and technically demanding, it can nevertheless provide a very informative profile which is robust and reproducible. Examples of such NMR-derived metabolic profiles from two individuals, one healthy and one with coronary heart disease, are shown in Figure 1. below.
Figure 1. Metabolic profiles of serum samples from a healthy individual (top) and a heart disease sufferer (bottom). Although grossly similar, subtle but reproducible differences (yellow insets) can be identified.
Although the profiles look similar at first glance, on closer inspection it is clear that there are subtle differences in the metabolic profile of different individuals. Unfortunately, there are also differences in the metabolic profile among healthy individuals (which might depend on all sorts of things, including recent diet, exercise and so forth). This makes it very difficult to spot any patterns (by visual inspection) that uniquely distinguish diseased and healthy individuals against all this background noise.
Fortunately, identifying subtle patterns in complex profiles is a task ideally suited to modern computers. Commercially available software exists which can analyze lists of numbers and identify any patterns within them. The pattern recognition software uses a list of numbers rather than a profile. The first step, therefore, is to convert the profile into such a list. Any real world object, such a metabolic spectrum, can be converted into a data vector in this way.
Next, the computer simplifies the complex data vector representing the profile using a process called dimensionality reduction. In simple terms, the computer looks to describe the complex profile using a much smaller list of numbers while still retaining the essential elements of the profile that distinguish healthy subjects from diseased subjects. Pronostics controls a number of proprietary methods for performing this dimensionality reduction on a metabolic profile that assist in the sensitive and specific diagnosis.
Application of these mathematical algorithms results in the reduction of the complex profile to just a handful of numbers describing each individual. These numbers are called Principal Components. The key step of educating our expert system involves teaching the computer exactly which Principal Components to extract from the profile. Once the Principal Components have been extracted, they can be plotted, giving a simple overview of how individuals differ in their profiles. Most importantly, the clustering of individuals with a particular disease is now very clear indeed (see the example in the next section). It is quite possible, using a plot such as this (which mathematicians call a Model) to predict the disease status of individuals from their molecular fingerprint alone.
Using Metabolic Profiles to Diagnose Coronary Artery Disease
In our proof-of-principle study, blood samples were prepared from almost 100 individuals, about half of whom had coronary heart disease, identified by invasive diagnostic angiography, while the remainder had normal coronary arteries. A metabolic profile was then obtained by NMR spectroscopy for each of these individuals, and the resulting profiles were analysed using the pattern recognition software. This reduced each profile to just two Principal Components which could be plotted on a simple graph. The results were astounding: on the basis of the metabolic profile alone, we were able to completely distinguish the heart disease sufferers from the healthy individuals.
This was the first time anyone had successfully applied a multivariate diagnostic technique to a prevalent disease. As a result, this study was reported in the leading scientific journal Nature Medicine in December 2002, and was carried by major national and international news media as a breakthrough in medical science.
The Search for an Economic High Throughput Profiling Solution
Since our pioneering report in 2002, a handful of other groups have reported successful application of multivariate diagnostic strategies to prevalent diseases. What these pioneering studies have in common, however, is the use of expensive, high technology approaches to generating the molecular profile. Worse still, the approaches themselves are public domain techniques developed over decades. It may be possible to find niche markets for NMR- or proteomic profile-based diagnostics, but the business case is usually borderline, the product always vulnerable to lower cost improvements and the scope for expansion strictly limited. Having demonstrated that multivariate diagnostics can and do work (and work very well), Pronostics therefore decided not to try and commercialise these expensive profiling methods,but instead to develop superior methods of obtaining useful biological fingerprints. Unlike all the existing examples, our new, methods had to be low cost, high throughput.
FingerPrint Immunomics
Finding a solution to the above problem required the invention of a whole new scientific discipline: immunomics. Like genomics (the study of genetic sequences), proteomics (the study of protein profiles) and metabolomics (the study of metabolic profiles), immunomics involves the generation of molecular profiles. For immunomics, that profile encodes information about the status of the individual’s immune system. Immunomic profiles can also be used to diagnose disease, but uniquely our methods of obtaining an immunological profile (FingerPrint technology) are robust, low cost and high throughput. Not only are our methods proprietary, but the whole discipline of immunomics is so new that Pronostics have filed broad claims to the very principle of a diagnostic test based on immunomics.
The principle of the test is to coat a library of antigens onto UltraPlex barcoded microparticles that allow the antigen to be subsequently identified easily. The barcoded library of antigens is then exposed to the blood sample from the patient, and antibodies bind to some of the antigens. These bound antibodies can then be detected using conventional fluorescent reporters. This process is outlined in Figure 2. below. While the principle of the test may be simple, the implementation requires very careful selection of the antigen library in order to obtain a diagnostically useful profile.
Fig 2.
There are an almost infinite number of possible antigens to include in the library, making efficient library design essential. Pronostics have developed several proprietary libraries which generate diagnostically useful Immunomic FingerPrints. It is this library design technology which is at the core of our proprietary FingerPrint technology platform.
It is clear from Figure 2. that generating an immunomic FingerPrint is a high-multiplex problem: we need to measure the binding of many different antibodies to different antigens in the library. An important step in the implementation of our FingerPrint technology is, therefore, its combination with Pronostics’ other platform technology, UltraPlex™. The properties of UltraPlex™ are ideally suited to generating low cost, robust and high throughput Immunomic FingerPrints. It was this obvious synergy between the technologies which underpinned the successful merger of SmartBead Technologies Ltd. (who developed UltraPlex™) and FingerPrint Diagnostics Ltd.
The World’s First Low Cost High-Throughput Multivariate Medical Diagnostic Product - CADprint™
Pronostics have configured a prototype product, based on one of our proprietary antigen libraries coated onto UltraPlex™ barcoded microparticles, and tested its ability to diagnose coronary heart disease using exactly the same samples used for our proof-of-principle studies with metabolic profiles. We called this prototype the CADprint™ test. Again, the results were stunning: CADprint™, even in its prototype format, could completely distinguish people with heart disease from people with normal coronary arteries on the basis of a single blood sample. Only this time (unlike the NMR-derived metabolic profiles), the test is low cost, and high throughput . CADprint™ is poised to become the World’s first low cost, high-throughput multivariate medical diagnostic product.
CADprint™ is now in full scale development with anticipated commercial launch during 2008 following extensive clinical trials.
UltraPlex™ FingerPrint Immunomics - A Platform Technology With Enormous Potential
UltraPlex™ FingerPrint Immunomics is a technology platform, not a single product.and Pronostics is confident of demonstrating its general applicability across a wide range of disease areas. Prior to the successful development of the CADprint™ prototype, Pronostics had demonstrated that the profiling approach could also diagnose other prevalent diseases with high sensitivity and specificity. For example, NMR-derived metabolic profiles can distinguish individuals with low bone mineral density (osteoporosis) from individuals with normal bone mineral density even more readily than they diagnose coronary heart disease sufferers.
Similar data have also been obtained for hypertension, diabetes and Alzheimer’s Disease. What these pilot studies clearly demonstrate is that multivariate diagnostics have the potential to revolutionise in vitro diagnostic provision across a broad range of the most prevalent disease areas. We do not presently know whether our immunomic tests can replicate this diagnostic power in these diseases as it did for coronary heart disease, but an important strand of our research activity over the next few years will be to obtain data using various antigen libraries in a number of carefully chosen disease areas where we believe such products would yield most patient benefit.
