Predicting Electrophoretic Mobility of Proteoforms for Large-Scale Top-Down Proteomics

Abstract

Large-scale top-down proteomics characterizes proteoforms in cells globally with high confidence and high throughput using reversed-phase liquid chromatography (RPLC)–tandem mass spectrometry (MS/MS) or capillary zone electrophoresis (CZE)–MS/MS. The false discovery rate (FDR) from the target–decoy database search is typically deployed to filter identified proteoforms to ensure high-confidence identifications (IDs). It has been demonstrated that the FDRs in top-down proteomics can be drastically underestimated. An alternative approach to the FDR can be useful for further evaluating the confidence of proteoform IDs after the database search. We argue that predicting retention/migration time of proteoforms from the RPLC/CZE separation accurately and comparing their predicted and experimental separation time could be a useful and practical approach. Based on our knowledge, there is still no report in the literature about predicting separation time of proteoforms using large top-down proteomics data sets. In this pilot study, for the first time, we evaluated various semiempirical models for predicting proteoforms’ electrophoretic mobility (μef) using large-scale top-down proteomics data sets from CZE–MS/MS. We achieved a linear correlation between experimental and predicted μef of E. coli proteoforms (R 2 = 0.98) with a simple semiempirical model, which utilizes the number of charges and molecular mass of each proteoform as the parameters. Our modeling data suggest that the complete unfolding of proteoforms during CZE separation benefits the prediction of their μef. Our results also indicate that N-terminal acetylation and phosphorylation both decrease the proteoforms’ charge by roughly one charge unit.

Evaluation of machine learning models for proteoform retention and migration time prediction in top-down mass spectrometry

“…A total of 7 proteoform features in three groups were tested for proteoform migration time prediction: the molecular mass and the charge state (group 1), the numbers of D, E, and N residues (group 2), and the numbers of L and I residues (group 3). The high accuracy of the semi-empirical model 38 shows that the two features in group 1 are important for migration time prediction. D, E, and N residues (features in group 2) slightly influence the proteoform charge, and L and I residues (group 3 features) have the highest hydrophobicity indexes in CZE experiments.…”

“…The semi-empirical model in ref. 38 predicted proteoform migration time in CZE-MS using the molecular mass M and charge Z of the proteoform. The molecular mass was included to predict the size of the proteoform.…”

“…Five machine learning models were assessed for predicting retention time in top-down RPLC-MS: the model in GPTime 24 , an FNN model, DeepRT+ 28 , Prosit 30 , and DeepDIA 32 . The last four models and a semi-empirical function 38 were also evaluated for predicting migration time in top-down CZE-MS. All the models were implemented in Python. The FNN and DeepRT+ models were implemented using the Pytorch package 42 , and the Prosit and DeepDIA models using the Keras package 43 with the TensorFlow backend.…”

“…The charge was estimated as the total number of positively charged amino acid residues (R, H, K, and the N-terminus) in the proteoform 13 . The electrophoretic mobility of the proteoform was predicted as µ = 𝑎 ln(1+0.35×𝑍) 𝑀 0.411 + 𝑏, where a and b are two parameters related to the CZE settings 38 . The electrophoretic mobility was then converted to its corresponding migration time using…”

“…Chen et al extended the semi-empirical model for peptide migration time prediction to proteoform migration time prediction in top-down MS 38 and obtained an R 2 = 0.98 on an E. coli CZE-MS data set. To the best of our knowledge, there have been no studies of the retention time prediction problem in top-down LC-MS.…”