Membrane curvature-sensing (MCS) proteins recognize and regulate the morphologies of biological membranes. As these proteins lack characteristic sequence motifs in their primary structure, they are not instantly recognizable by genomic databases. Overcoming this technological challenge toward the agile identification of new proteins can promote the elucidation of membrane morphological regulation. Here, for the selective identification of MCS proteins, comparative proteomic analysis was performed using different sizes of the spherical supported lipid bilayer (SSLB), which consists of spherical SiO2 particles covered with a lipid bilayer. Because of the presence of SiO2 core, the curvature of the surrounding membrane is well-controlled and stable even on a micron scale. To prove this concept, known membrane curvature-sensing protein domains, Bin/Amphiphysin/Rvs (BAR) and Epsin N-terminal homology (ENTH), were evaluated by performing a binding assay using SSLBs, and the preferential binding to the highly curved membrane was confirmed. Peripheral membrane proteins obtained from normal human dermal fibroblast (NHDF) and human breast cancer (MDA-MB-231) cells were used in shotgun proteomic analysis, and 786 and 949 proteins were identified from SSLBs as lipid membrane binders, respectively. Statistical quantitative analyses of proteins detected from each SSLB with a different size revealed 118 candidate proteins, including 23 proteins unique to MDA-MB-231 cells, as membrane curvature sensors, including some previously reported curvature sensors. Functional clustering analysis based on the KEGG orthology database revealed that the protein-binding property to specific high or low membrane curvature correlated with their functions. Further investigation of candidate proteins will lead to the identification of new MCS proteins as well as cancer biomarkers.