← Back to home

DART-ID Increases Single Cell Proteome Coverage

DART-ID is a computational method which can increase the coverage of LC-MS/MS proteomics experiments – specifically those relying on MS2-based quantitation (i.e., isobaric tag labeling, TMT). It uses inferred and observed peptide retention time (RT) to update the peptide identification confidence, and usually results in up to 50% more peptides observed at 1% FDR.

The paper is currently under review at PLOS Computational Biology.

In addition to the accompanying posters, this Twitter thread is a good introduction to the project:


As first author, I lead the development of most of the project shortly after its conception. Working with my collaborator Alex Franks from UCSB, I first explored various iterations of the method on our SCOPE-MS method development data. The first version of DART-ID used pairwise alignment between experiments, and chose a reference experiment based on a couple of heuristics. Alex implemented a global alignment method in the Stan modeling language that greatly improved our performance and we iterated on that method from then on out.

After tweaking our RT alignment model to a good place, I decided to port the DART-ID tool from R to python to increase its usability and speed. (the R version in the DART-ID_2018 repo is still functional but it lacks some key features).

I wrote the original drafts of the manuscript, conducted the analyses, and created the figures/visualizations shown in the paper. I also lead the work of responding to our first round of reviews. Hopefully PLOS will start publishing these reviews openly…

Of course – thanks to my co-authors Alex Franks and Nikolai Slavov, as well as members of the Slavov Lab for their input and feedback. It truly would not have been possible without their help.

← Back to home