Biom-60. apollo: raman-based pathology of malignant glioma

Methylation classification is an essential component for integrative diagnosis in glioma, however, the DNA methylation classification is not always available for all the samples. We hypothesized that Raman spectroscopy might be suitable to predict the glioma methylome, based upon its ability to create a molecular fingerprint of the tumor and would provide biological insights into the discriminatory features. Raman Spectroscopy was used for molecular fingerprinting of the regions of interest using 1mm2 FFPE tissue spots from 45 patient samples with LGm1 to LGm6 methylation subtypes. Spectral information was then used to train a convolutional neural network (CNN), capable of detecting the glioma methylation subtypes. 70 % of the dataset was used for model training while the remaining 30% for validation. We demonstrate that Raman spectroscopy can accurately and rapidly classify gliomas according to their methylation subtype from achieved FFPE samples, as a novel way to obtain classification. For each sample we ran Ward linkage clustering with a variable number of clusters (from 2 to 7), with the majority cluster corresponding to tumor spots and the others corresponding to (various types of) non-tumor spots. The average accuracy over all samples was 90:3%, the average precision was 99:6% and the average recall was 90:2%. We show that Raman spectroscopy together with artificial intelligence can predict the methylome of glioma samples and augment the ability to classify these tumors retrospectively. The non-destructive nature of this method and the ability to be applied on FFPE samples directly, allows the histopathologist to reuse of the same slide for subsequent staining and downstream analyses.