In recent years machine learning has made extensive progress in modeling many aspects of mass spectrometry data. We brought together proteomics data generators, repository managers, and machine... Show moreIn recent years machine learning has made extensive progress in modeling many aspects of mass spectrometry data. We brought together proteomics data generators, repository managers, and machine learning experts in a workshop with the goals to evaluate and explore machine learning applications for realistic modeling of data from multidimensional mass spectrometry-based proteomics analysis of any sample or organism. Following this sample-to-data roadmap helped identify knowledge gaps and define needs. Being able to generate bespoke and realistic synthetic data has legitimate and important uses in system suitability, method development, and algorithm benchmarking, while also posing critical ethical questions. The interdisciplinary nature of the workshop informed discussions of what is currently possible and future opportunities and challenges. In the following perspective we summarize these discussions in the hope of conveying our excitement about the potential of machine learning in proteomics and to inspire future research. Show less
While mass spectrometry still dominates proteomics research, alternative and potentially disruptive, next-generation technologies are receiving increased investment and attention. Most of these... Show moreWhile mass spectrometry still dominates proteomics research, alternative and potentially disruptive, next-generation technologies are receiving increased investment and attention. Most of these technologies aim at the sequencing of single peptide or protein molecules, typically labeling or otherwise distinguishing a subset of the proteinogenic amino acids. This note considers some theoretical aspects of these future technologies from a bottom-up proteomics viewpoint, including the ability to uniquely identify human proteins as a function of which and how many amino acids can be read, enzymatic efficiency, and the maximum read length. This is done through simulations under ideal and non-ideal conditions to set benchmarks for what may be achievable with future single-molecule sequencing technology. The simulations reveal, among other observations, that the best choice of reading N amino acids performs similarly to the average choice of N+1 amino acids, and that the discrimination power of the amino acids scales with their frequency in the proteome. The simulations are agnostic with respect to the next-generation proteomics platform, and the results and conclusions should therefore be applicable to any single-molecule partial peptide sequencing technology. Show less
