Spatially resolved transcriptomics and graph-based deep learning improve accuracy of routine CNS tumor diagnostics.

Area of Science:

• Neuro-oncology
• Molecular pathology
• Computational biology

Background:

• Brain tumor diagnosis relies on histopathology and molecular markers like DNA methylation and NGS.
• Limited DNA quantity and quality restrict the feasibility of conventional molecular diagnostics.
• Accurate tumor classification is crucial for effective treatment and patient outcomes.

Purpose of the Study:

• To introduce NePSTA (neuropathology spatial transcriptomic analysis) for integrated morphological and molecular neuropathological diagnostics.
• To assess NePSTA's accuracy in classifying central nervous system (CNS) tumor histology and molecular subtypes.
• To demonstrate NePSTA's utility in challenging samples with minimal DNA input.

Main Methods:

• NePSTA utilizes spatial transcriptomics and graph neural networks for automated analysis of 5-µm tissue sections.
• The method was trained and evaluated on 130 participants with CNS malignancies and healthy donors from four medical centers.
• NePSTA reconstructs immunohistochemistry and genotype profiles from minimal tissue requirements.

Main Results:

• NePSTA achieved high accuracy in predicting tissue histology and methylation-based CNS tumor subclasses.
• The technique successfully performed molecular profiling on samples with insufficient DNA for conventional methods.
• Demonstrated potential for enhanced tumor subtype identification and precise diagnostic workup.

Conclusions:

• NePSTA offers a novel approach for comprehensive neuropathological diagnostics from limited tissue samples.
• This method overcomes DNA quantity limitations, enabling advanced molecular classification.
• NePSTA has significant implications for improving the speed and precision of brain tumor diagnosis and subtyping.