AI may advance cancer detection through glycan analysis

A study published in Nature Methods suggests that a new deep-learning model can detect cancer through glycan analyses more rapidly and efficiently than the current semi-manual method.

Training the dilated residual neural network has enabled the AI model, dubbed CandyCrunch, to “calculate the exact sugar structure in a sample in 90% of cases," said researcher Daniel Bojar. "We believe that glycan analyses will become a bigger part of biological and clinical research now that we have automated the biggest bottleneck."

The model was trained on a database of over 500,000 examples of different fragmentations and associated structures of glycan molecules. Researchers said it could predict glycan structure "in seconds" from raw liquid chromatography tandem mass spectrometry data, with accuracy as high as 90.3%. 

Moreover, it could identify structures that are often missed by human analyses due to their low concentrations. The authors suggested that CandyCrunch may accelerate the discovery of glycan-based biomarkers for the diagnosis and prognosis of cancer.