Low Abundance Peptide Quant by Mining MS1 PTM Variants

Was this article helpful?

View our ASBMB/Molecular and Cellular Proteomics 2019 poster HERE 

In tandem mass spectrometry (MSMS), 95%+ of the information resides in MS1 m/z data because relatively few precursors get fragmented. Yet conventional data analysis focuses primarily on MS2 spectra and sequence statistics, with MS1 data treated as an afterthought. This backward approach only scratches the surface of what’s possible.

“Deep” label-free quantitation (DLFQ) instead focuses on identifying/quantifying MS1 precursor ions using MS2-identified peptides as a starting guide. This is like interpreting Egyptian hieroglyphs as an integrated story, instead of a collection of unconnected sentences, guided by a Rosetta Stone. It allows significantly more peptides and proteins to be characterized, particularly those of low abundance or with post-translational modifications (PTM) of clinical relevance.

Instead of looking at a multi-hour MSMS run as a single experiment, deeper analysis is possible by analyzing it as thousands of one-second mini-experiments. Here, every m/z data-point is standalone evidence of an ion even if it appears only once or twice.

For low abundance peptides, thousands of anonymous ones are manifested as a single pair of isotopic MS1 data-points. That’s enough to determine charge, precursor mass, retention time, and rough quantity (i.e. apex intensity). But without a way to infer sequence, they are typically discarded as noise. We show how powerful IT can uncover mass interrelationships that imply sequence information.

Co-eluded peptides confuse ion assignments and cause irreproducible quantitation. AI can in principle combat this, but most techniques require extensive annotated datasets that proteomics lacks. We discovered that using an AI image processing technique, computer vision (CV), can be applied without annotated training data to the extent that data can be interpreted as an image. We show how CV’s Hough Transform for finding lines can detect co-eluded peptides.

DLFQ can uniquely identify and quantify low-abundance, modified peptides otherwise out of reach.