High-dimensional spectroscopy often provides rich information. Raman spectroscopy is a non-destructive molecular sensing method. However, Raman signals in bio-samples are hard to interpret, due to the high dimensionality. Measurement of optical spectroscopy is also complicated and requires high-end instrumentations and intricate data analysis techniques. Machine learning methods offer great opportunities to extract subtle and deep information in high-dimensional spectra. They can also assist measurement of complex optical spectroscopy of materials with simpler optical setups. In this work, we develop a platform that enables rapid screening of AD biomarkers by employing graphene-assisted Raman spectroscopy and machine learning interpretation in animal brains. The method facilitates the study of AD and can be extended to other tissues, biofluids, and for various other diseases. We also propose a computational reflectometry approach based on a deep learning model called ReflectoNet. It predicts complex refractive indices of thin films on top of nontrivial substrates from reflectance spectra, which was not feasible previously.