Molekular-gesteuerte Neurochirurgie: Bestimmung des Neural-Scores durch intraoperative epigenetische Profilerstellung

The extent of integration of glioblastoma cells into the brains neural circuit has an important role in the tumour progression. The more integrated glioblastoma requires a larger extent of resection. Intraoperative molecular profiling of this "neural" subtype remains challenging. Here, we demonstrate a novel algorithm to accurately detect the epigenetic neural phenotype intraoperatively within less than an hour.

In this study, 23 patients were enrolled and 1-3 biopsies were multiplexed and barcoded for intraoperative sequencing. DNA extraction and library preparation was performed using the rapid-UKER protocol. Sequencing was performed by MinION system on the R10.4 Flow cells. Prediction of the neural phenotype was performed by a contrastive-based deep-learning model using cross-attention for accurate prediction.

Mean DNA extraction and library preparation (tissue2sequencing) time was 48.76 minutes using the rapid-UKER protocol containing transposase adapters technology. Mean fragment length of the DNA was 5428 +/- 1432 bp. Sequencing speed was between 953-2132 CpG sites per minute resulting in accurate predictions after 10 minutes (two sequencing fasa5 batches) in reality. Computational simulation demonstrate that the mean prediction time was 7.23 +/- 2.13 minutes.

Intraoperative epigenetic profiling is an accurate tool to predict the neural phenotype and may be used to molecular-guided resection in the future.