One of the greatest challenges for mass spectrometry (MS)-based proteomics is the limited ability to analyze small samples. Here we investigate the relative contributions of liquid chromatography (LC), MS instrumentation and data analysis methods with the aim of improving proteome coverage for sample sizes ranging from 0.5 ng to 50 ng. We show that the LC separations utilizing 30-mum-i.d. columns increase signal intensity by >3-fold relative to those using 75-mum-i.d. columns, leading to 32% increase in peptide identifications. The Orbitrap Fusion Lumos MS significantly boosted both sensitivity and sequencing speed relative to earlier generation Orbitraps (e.g., LTQ-Orbitrap), leading to a ~3-fold increase in peptide identifications and 1.7-fold increase in identified protein groups for 2 ng tryptic digests of the bacterium S. oneidensis. The Match Between Runs algorithm of open-source MaxQuant software further increased proteome coverage by ~ 95% for 0.5 ng samples and by ~42% for 2 ng samples. Using the best combination of the above variables, we were able to identify >3,000 proteins from 10 ng tryptic digests from both HeLa and THP-1 mammalian cell lines. We also identified >950 proteins from subnanogram archaeal/bacterial cocultures. The present ultrasensitive LC-MS platform achieves a level of proteome coverage not previously realized for ultra-small sample loadings, and is expected to facilitate the analysis of subnanogram samples, including single mammalian cells.