Aqsens Health has developed a novel fuzzy assay platform that can point out diseases and conditions early on. In a new publication in Bioanalytical Chemistry by Kulpakko et al, this Aqsens Health’s unique assay platform has been utilized in detection of urinary tract infection and it’s causes. The unique combination of the proprietary and sensitive E-TRF method and a semi-specific probe together with machine learning shows unique promise in fast diagnosis of multicity of diseases. In disease, the diagnosis can often be reached by multiple means – take Influenza: one could have a specific test for the particular influenza-virus – or combine the symptoms of the patient and end up with the same conclusion. It is known that people with influenza usually have e.g. high fever, aching muscles and sore throat. Now we all know that aching muscles could result also from heavy exercise but combined with high fever and other symptoms of Influenza the likelihood for correct diagnosis increases tremendously. In many cases there are no specific quick tests for a particular disease, but the diagnosis is mostly symptom based; and this often works remarkably well – when the symptoms are clear and cannot be mixed with other diseases.
Unfortunately, this confusion, or even lack of symptoms prevents efficient and early diagnosis of many and even severe conditions. The same applies to chemical marker based specific tests – many life-threatening conditions are not possible to diagnose with individual chemical marker-based tests but require laborious and costly diagnostic approaches and skillful medicine. These complicated tests cannot be applied as screening tools, especially if special expertise is needed – thus in many cases, the diagnosis is delayed until later stages of diseases, sometimes even fatally.
Aqsens Health’s E-TRF approach is chemical, but it is not marker based in the traditional sense. In the core of the technology lies artificial intelligence – a machine learning based algorithm that groups the E-TRF chemical assay-based signals into two groups: healthy, disease X. In the development phase of the algorithm, the E-TRF array from patient samples is chemically tuned to recognize differences between the representative patient groups. As the algorithm has learned to differentiate between the groups with good sensitivity and specificity, the E-TRF array parameters are fixed. After this multidimensional learning process the E-TRF array – algorithm pair can be used to screen for disease X. Because the chemical tuning is aimed for separation of the two groups of patients, even the molecular species recognition capacity of each of the E-TRF array chemistries needs not to be known – the chemistries can be anything from nonspecific to specific whereas the array itself can be seen as a fuzzy “black-box” that delivers the requested result to the machine learned algorithm that itself is a product of a fuzzy learning process. This very method was used in the Kulpakko et al publication with an exception in disease grouping – the fuzzy process was used to spot a very specific target Escherichia coli, a pathogen that should not be present in our urine.