We propose an updated approach for approximating the isotope distribution of average peptides given their monoisotopic mass. Our methodology involves in-silico cleavage of the entire UNIPROT database of Human reviewed proteins using Trypsin, generating a theoretical peptide dataset. The isotope distribution is computed using BRAIN. We apply a compositional data modelling strategy that utilizes an additive log-ratio transformation for the isotope probabilities followed by a penalized spline regression. Furthermore, due to the impact of the number of Sulphur atoms on the course of the isotope distribution, we develop separate models for peptides containing zero up to five Sulphur atoms. Additionally, we propose three methods to estimate the number of Sulphur atoms based on an observed isotope distribution. The performance of the spline models and the Sulphur prediction approaches is evaluated using a mean squared error and a modified Pearson’s χ² goodness-of-fit measure on an experimental UPS2 data set. Our analysis reveals that the variability in spectral accuracy contributes more to the errors than the approximation of the theoretical isotope distribution by our proposed average peptide model. Moreover, we find that the accuracy of predicting the number of Sulphur atoms based on the observed isotope distribution is limited by measurement accuracy.

The continuous advancements in LC-MS/MS proteomics over the past decades have paved the way for transformative changes in the field of medicine, particularly in the realms of preventive and personalized healthcare. Many new algorithms are evaluated on unknown proteomes and using databases with annotated MS2-spectra. When the research is focused on MS1-spectra, such databases are not available yet. Specifically, we propose a comprehensive workflow to extract MS1 isotope distributions from spectra, which we validated using a proteomics standard kit comprising known proteins at varying concentrations in duplicate. Our workflow incorporated a database search utilizing a state-of-the-art algorithm at 1% FDR. Through this approach, we investigated the impact of protein concentration on the probability of protein identification. Confidently identified PSMs were used to extract the MS1 isotope distributions through the proposed workflow. A total of 138.111 MS1 isotope distributions were extracted. Isotope distributions with 2 or more peaks were compared with their theoretical isotope distributions using the spectral angle. A median spectral angle of 0,101 and 0,0992 was observed in both samples indicating a high similarity. The findings from this study were compiled into a dataset which can potentially facilitate the development of novel tools with a focus on MS1 data.