In neurodegenerative diseases, proteins fold into amyloid structures with distinctconformations (strains) that are characteristic of different diseases. However, thereis a need to rapidly identify amyloid conformations in situ. Here, we use machinelearning on the full information available in fluorescent excitation/emission spec-tra of amyloid-binding dyes to identify six distinct different conformational strainsin vitro, as well as amyloid-β (Aβ) deposits in different transgenic mouse models. OurEMBER (excitation multiplexed bright emission recording) imaging method rapidlyidentifies conformational differences in Aβ and tau deposits from Down syndrome,sporadic and familial Alzheimer’s disease human brain slices. EMBER has in situidentified distinct conformational strains of tau inclusions in astrocytes, oligoden-drocytes, and neurons from Pick’s disease. In future studies, EMBER should enablehigh-throughput measurements of the fidelity of strain transmission in cellular andanimal neurodegenerative diseases models, time course of amyloid strain propagation,and identification of pathogenic versus benign strains.

Custom developed MATLAB code for micrograph particle segmentation and EMBER analysis as well as raw and normalized in vitro fibril EMBER data.