Machine Learning Boosted Quantum-Profiling of Blood Antigens

Area of Science:

• Glycosciences
• Computational Chemistry
• Bioinformatics

Background:

• Carbohydrate characterization is vital for glycosciences but hindered by structural complexity.
• Existing methods like NMR and mass spectrometry face limitations with complex saccharides and regioisomeric linkages.

Purpose of the Study:

• To develop a computational methodology for simultaneous recognition of diverse blood antigens.
• To overcome limitations of conventional techniques in carbohydrate structural analysis.

Main Methods:

• Utilized a quantum tunneling method combined with machine learning (ML).
• Employed a random forest classifier with SHapley Additive exPlanations (SHAP) for interpretability.
• Performed rapid quantum profiling of molecules based on transmission signatures.

Main Results:

• Achieved simultaneous recognition of a wide range of blood antigens with good precision and sensitivity.
• Demonstrated the capability for accurate and high-throughput "sugar calling" of carbohydrates.
• Provided interpretable results using SHAP for the ML model.

Conclusions:

• The ML-enhanced quantum methodology offers a powerful alternative to traditional carbohydrate analysis.
• Facilitates accurate and high-throughput characterization of complex carbohydrates.
• Enables efficient identification of blood antigens through their transmission signatures.