Machine learning identifies experimental brain metastasis subtypes based on their influence on neural circuits.

Area of Science:

• Neuroscience
• Oncology
• Computational Biology

Background:

• Brain metastases frequently cause neurocognitive symptoms.
• The mechanisms by which tumors affect neuronal circuits beyond mass effect are poorly understood.

Purpose of the Study:

• To comprehensively model brain functional alterations in brain metastasis.
• To identify molecular and functional changes underlying neurocognitive deficits in brain metastasis.

Main Methods:

• Utilized diverse preclinical models of brain metastasis with varying primary sources and oncogenic profiles.
• Performed multidimensional brain functional analyses, including local field potential oscillations.
• Analyzed transcriptomic and mutational profiles.
• Applied machine learning strategies to brain activity readouts.

Main Results:

• Dissociated heterogeneous impacts on cortical and hippocampal oscillatory activity from homogeneous tumor size or glial response.
• Identified model-specific molecular programs impairing neuronal crosstalk.
• Confirmed model-specific brain activity alterations using machine learning, enabling prediction of metastasis presence and subtype.

Conclusions:

• Brain metastasis disrupts neuronal function through mechanisms beyond physical mass effect.
• Specific molecular programs and brain activity patterns characterize different brain metastasis subtypes.
• Machine learning can predict brain metastasis characteristics from functional readouts.