Vesicles are small ”bubbles" in the size of viruses and are released by all cells, including immune cells. Via the blood they are transported throughout the body, reaching every organ and cell system to induce specific biochemical processes.
The immune system uses these vesicles mainly to instruct and activate other immune cells. This function is conveyed through enzymes (proteases) and other, non-enzymatic factors. The latter include chemokines, cytokines, and mRNA, among others. The combination of molecules present in vesicles differs depending on their designed purpose. This again is adjusted to the type and severity of the disease process that triggered the immune response in the first place. The decoding of the factor composition gives a new and time-dependent insight into the function of the immune system and the disease that was targeted.
The diagnostic process starts with a blood sample at your local doctor. The first screening analysis is performed using a microfluidic device (V-Disk Player) at the point of care. The obtained data are transferred to the cloud where they are analyzed by AI Algorithms. The result are returned to the customers mobile device and/or the doctor's office (computer).
In case the result indicates a potential medical condition, the physician will discuss the result with his patient and have a confirmation analysis performed in a central laboratory. If this test confirms the screening result (e.g. a suspected cancer), the patient is referred for further clinical examinations (e.g. imaging).
Blood plasma is a complex soup of free proteins and particles, such as lipoproteins, of which extracellular vesicles (EVs) are only a small minority. To separate out EVs, we make use of dual-mode chromatography (DMC), which is a patented method developed by one of our team members.
In DMC, the plasma sample is loaded on top of a column containing two layers of different types of beads. As the sample migrates through the DMC column, it will first encounter a 'size-exclusion' layer, consisting of beads with ca. 40 nanometer (nm) pores. Free proteins and smaller particles, such as small lipoproteins, will enter the pores and be delayed. Particles that are larger than 40 nm, such as EVs and large lipoproteins, will migrate faster through this layer. To separate those two remaining components, we make use of an 'ion exchange' layer, consisting of negatively-charged beads. The large lipoproteins, which are positively charged, will be captured by those beads, while negatively-charged EVs are repelled and will migrate through this layer unchallenged. The end result of the DMC procedure is a highly EV-enriched sample.
As DMC is fast, reproducible and automatable, it is very suitable for use in a clinical research setting and central laboratories.
To measure the activity of different proteases in plasma EV samples, we make use of a fluorescence-based assay that we developed and patented.
First, we design peptides with a specific amino acid sequence that corresponds to the recognition site used by a given protease when cutting its substrates. At one end of the peptide, a fluorescent molecule is attached, while a quencher is attached to the other end. An intact peptide will not emit any signal. Upon addition of these peptides to the EV sample, and subsequent cleavage by their specific protease, the fluorophore is released from the quencher and a signal will be emitted. Recording the increase of this signal over time is used as a measure for the amount of active protease that is present.
By using different peptides designed to be cleaved by different proteases, we can obtain a protease activity pattern that is unique for each sample. These patterns for different patient populations are then analyzed by artificial intelligence to build classification algorithms.
The analysis of multiple molecular features in plasma EVs, like protease activity, can be used to detect characteristic patterns in patients for diagnostic purposes. However, when profiling dozens or hundreds of molecules, finding these patterns is only possible with the help of artificial intelligence-based computational models (AI models).
In KI-VesD, we employ advanced bioinformatics and statistical methods to select EV protease activities that characterize specific groups of cancer patients. Next, we use supervised machine learning to train AI models with data obtained from hundreds of patients and obtain predictive classification models for melanoma, prostate cancer, lymphoma and other tumor entities.
Our vision is that these AI models can be used to accurately predict the risk of tumor relapse in patients or correctly provide an early diagnosis, e.g. of a developing prostate carcinoma.
In addition, our AI-based computer models can be integrated in cloud-based IT solutions that enable to provide quick diagnostics results.
The automation of diagnostic tests increases the robustness and reliability of the diagnosis and minimises time and costs. In the KI-VesD project, Hahn-Schickard is developing a microfluidic cartridge (V-Disk) and devices to automate the test. The patient's blood sample is placed on the CD-shaped cartridge. Vesicles are then extracted from the sample in the "disk player" using microfluidics and analysed with the protease activity test. The activity patterns are pseudonymised and transmitted in encrypted form to a server for pattern recognition.
The V-Disk cartridge is a microfluidic chip that can be used to automate biochemical processes. It is, so to speak, a miniaturized laboratory ("lab-on-a-chip"), which makes it possible to extract the vesicles from the blood sample by means of various filtration and chromatography steps without further intervention.
