Regional
Novel tool in immunohematology
Morten Hanefeld Dziegiel
Copenhagen University Hospital, Denmark
Anders Kristensen
DTU Health Tech, Denmark
Emil Alstrup Jensen
Technical University of Denmark, Denmark
Raman spectroscopy used for blood grouping and other molecular phenotyping enabling personalized transfusion medicine.
Background
Raman spectroscopy is a method for molecular typing based on specific and reproducible interaction of light (photons) and matter (vibrations in molecules in the blood sample).
Monochromatic laser-light hits the blood sample and will be scattered into distinct colors deviating from the color of the incoming laser. The exact collection of colors, the spectrum, of scattered light is determined by the exact composition of the blood and the exact isoforms of individual molecules. The color of the scattered light is recorded by a Raman spectroscope. In practice most molecules in a biological sample (e.g., proteins, DNA, RNA, lipids, carbohydrates, hydrocarbons (alkanes, alkenes, alkynes, aromatic compounds), polymers (Polyethylene, polystyrene, polyvinyl chloride (PVC) and more) will be identifiable by their exact Raman scattering of incoming laser light. The number of informative peaks identified in this work on ABO determination was 24 peaks of extraordinary importance. The physical background for Raman scattering is photons interacting with vibrational modes of the chemical bonds between atoms.
The isotyping is done with crude whole blood, no sample preparation, no use of labeling reagents. The complexity of Raman spectra of whole blood has previously hindered the use of Raman spectra for blood grouping. However, the advent of artificial intelligence and corresponding computing power have enabled extracting the correlation between specific spectral properties and reference data on blood groups and phenotypes, and thus the deduction of the blood groups from the Raman spectrum.
The aim of this study (1) was to explore the use of Raman spectroscopy and artificial intelligence for determination of properties of molecules of relevance for personalized transfusion medicine.
Methods
We examined EDTA whole blood from voluntary blood donors. We used an inverted microscope with an integrated Raman spectroscope and a 785 nm laser as the light source. The recording of the spectrum was done of whole blood flowing past the laser beam in a capillary flow cell. Raman scattered photons were analyzed in a spectrometer and recorded by a camera. The examination time of each sample was 60 seconds.
Results
Our training set of donor samples was deliberately assembled to ensure approximately equal numbers of blood samples from donors with blood groups A, B, O and AB donors (total of 270 donors). We compared the six combinations to describe informative spectral properties. The ability to distinguish the different ABO blood groups was tested and described by metrics like area under curve (AUC) for the Receiver Operating Curve (ROC), accuracy, sensitivity, specificity. We obtained a mean of 0.9 for the AUC of the six ABO combinations. We then examined the ability to predict additional phenotypes from the 51 allotypes determined by our standard donor genotyping (2, 3) and phenotyping. From the same spectral dataset used for the ABO determination, a total of 32 additional properties could be deduced with a mean accuracy of 74%. The remainder of the allotypes were not discernable, presumably due to skewed frequencies of representation of the isoforms. Of special interest to the immunohematologist was the finding of an AUC-ROC of 80 % of donors of blood group A with anti-B above or below a titer of 10.
Summary/conclusions
A combined process with label-free Raman spectroscopy of non-processed whole EDTA blood samples, data analysis by artificial intelligence using reference information of blood groups and phenotypes, demonstrates a promising proof-of-principle for fast and comprehensive donor isotyping. The performance level of the method in its present form is not ready for clinical use.
References
1. Jensen, E. A., Serhatlioglu, M., Uyanik, C., Hansen, A. T., Puthusserypady, S., Dziegiel, M. H., & Kristensen, A. (2024). Label-Free Blood Typing by Raman Spectroscopy and Artificial Intelligence. Advanced Materials Technologies, 9(2), 1-16. Artikel 2301462. https://doi.org/10.1002/admt.202301462
2. Krog GR, Rieneck K, Clausen FB, Steffensen R, Dziegiel MH. Blood group genotyping of blood donors: validation of a highly accurate routine method. Transfusion. 2019 Oct;59(10):3264-3274. doi: 10.1111/trf.15474. Epub 2019 Aug 15. PMID: 31415105.
3. Johnson L, Coorey CP, Marks DC. The hemostatic activity of cryopreserved platelets is mediated by phosphatidylserine-expressing platelets and platelet microparticles. Transfusion. 2014;54:1917-26.
