Proton (1H) Magnetic Resonance Spectroscopy (MRS) is a powerful tool for studying brain metabolites in-vivo. As a clinical research tool, consistent and accurate quantification is a necessity, which has driven the MRS community to prioritize producing reliable and standardized analysis software packages. Though, low signal to noise, artifacts, and overlapping signals still provide significant challenges, especially for deep brain regions and multi-voxel acquisition. Data quality can be improved through collecting multiple transients or employing multi-scan (editing, nulling, etc.) methods, but time constraints may limit what can be done during a scan session. This thesis aims to address some of the fundamental problems associated with performing MRS in the brain by developing advanced methods surrounding MRS data processing and analysis. Specifically, I’ll describe the results accomplishments from 2 projects. In part-1, I describe developing the open-source COHERENC database and meta-analysis to provide a metabolic profile in both the healthy and clinical brain. Then, in Part-2, I describe the development of a neural network training and benchmark dataset (AGNOSTIC) for MRS as well as our work in creating deep-learning-based data analysis techniques.