Improve the quantitative analysis of post-translationally modified peptides by dedicated statistical modeling

[veitveit]

Abstract

The characterization of post-translational modifications (PTMs) by bottom-up mass spectrometry is crucial to describe cellular processes. However, current approaches are hampered by the lack of suitable statistical workflows to precisely determine relative PTM changes. Compared to protein quantification where multiple peptide measurements are available per protein and erroneous measurements can be discarded or down-weighted, PTM quantification relies merely on one or few peptide-spectrum matches. Additionally, one needs to distinguish changes of protein quantity from PTM modulation by correcting for the differential protein abundance. This doesn’t account for differences of variability of modified peptides and proteins nor the uncertainty of the relative protein abundance estimate. Finally, although peptide-spectrum matches corresponding to the same PTM often present distinct charge states, miscleavages and combinatorial modification, there is no agreement in the field on how to group them before statistical analysis.
We aim to develop/implement an ion-centric framework to quantify protein modifications while correctly accounting for the different levels of variability. Simultaneously modelling all peptides of a protein will allow a direct estimation of the PTM regulation that is adjusted to differential protein abundance and accounts for the estimated uncertainty.
The results will be presented as guidelines on the proteomics-academy.org website and through publication.

Work plan

• Discussion of featured data types and selection of a set of datasets suitable for testing and benchmarking
• Define general guidelines for the analysis of PTMomics data
• Outline statistical frameworks for ion-centric inference
• Implementation of ion-centric models
• Evaluate the approaches on a quantitative benchmark set
• Summarize guidelines and results

Benefits

• Participants will gain expertise in the state-of-the-art models for protein quantification and have influence on future data analysis standards
• The project will provide guidelines for the analysis of modified peptides, an essential step for making PTMomics data more reproducible and comparable.

Technical details

Programming language is preferably R
Existing workflows for protein-level inference will be adapted towards ion-centric inference.
A suitable example data set with ground truth (http://www.nature.com/nbt/journal/v35/n8/full/nbt.3908.html) and selected public and in-house data sets will be available.

Contact information

Veit Schwämmle
Department of Biochemistry and Molecular Biology
University of Southern Denmark
Denmark
[email protected]

Lieven Clement
Department of Applied Mathematics, Computer Science and Statistics
Ghent University
Belgium
[email protected]

Marie Locard-Paulet
National Center for Scientific Research
Institute of Pharmacology and Structural Biology
France
[email protected]