A learned score function improves the power of mass spectrometry database search

Abstract

Motivation: One of the core problems in the analysis of protein tandem mass spectrometry data is the peptide assignment problem: determining, for each observed spectrum, the peptide sequence that was responsible for generating the spectrum. Two primary classes of methods are used to solve this problem: database search and de novo peptide sequencing. State-of-the-art methods for de novo sequencing use machine learning methods, whereas most database search engines use hand-designed score functions to evaluate the quality of a match between an observed spectrum and a candidate peptide from the database. We hypothesized that machine learning models for de novo sequencing implicitly learn a score function that captures the relationship between peptides and spectra, and thus may be re-purposed as a score function for database search. Because this score function is trained from massive amounts of mass spectrometry data, it could potentially outperform existing, hand-designed database search tools.

Results: To test this hypothesis, we re-engineered Casanovo, which has been shown to provide state-of-the-art de novo sequencing capabilities, to assign scores to given peptide-spectrum pairs. We then evaluated the statistical power of this Casanovo score function, Casanovo-DB, to detect peptides on a benchmark of three mass spectrometry runs from three different species. In addition, we show that re-scoring with the Percolator post-processor benefits Casanovo-DB more than other score functions, further increasing the number of detected peptides.

Figure 1.

Each figure plots the number of peptides detected as a function of FDR threshold for the (a) E.coli, (b) yeast, and (c) human datasets. In each plot, the series correspond to different score functions, and the 1% FDR threshold is highlighted with a red dashed line. (d) Similar to panel (c), but generated using our in-house search engine.

Figure 2.

Each figure plots the number of peptides detected after Percolator post-processing as a function of FDR threshold for the (a) E.coli, (b) yeast, and (c) human datasets. In each plot, the series correspond to different score functions, and the 1% FDR threshold is highlighted with a red dashed line.

Figure 3.

An upset plot showing the overlap in peptide detections at 1% FDR between Casanovo-DB, Tide, SAGE, and MaxQuant on the human dataset.

Figure 4.

Each figure plots the number of peptides detected as a function of FDR threshold for the E.coli dataset, broken down by whether the precursor had an m/z in (a) the bottom quartile range of 350–467 m/z or (b) the top quartile range of 704–1389 m/z. In each plot, the series correspond to different score functions, and the 1% FDR threshold is highlighted with a red dashed line. Casanovo-DB performs much worse on low m/z precursors compared to those with high m/z, exemplifying the calibration problems which are resolved by Percolator.