A Matter of Time: Faster Percolator Analysis via Efficient SVM Learning for Large-Scale Proteomics

Abstract

Percolator is an important tool for greatly improving the results of a database search and subsequent downstream analysis. Using support vector machines (SVMs), Percolator recalibrates peptide-spectrum matches based on the learned decision boundary between targets and decoys. To improve analysis time for large-scale data sets, we update Percolator's SVM learning engine through software and algorithmic optimizations rather than heuristic approaches that necessitate the careful study of their impact on learned parameters across different search settings and data sets. We show that by optimizing Percolator's original learning algorithm, l-SVM-MFN, large-scale SVM learning requires nearly only a third of the original runtime. Furthermore, we show that by employing the widely used Trust Region Newton (TRON) algorithm instead of l-SVM-MFN, large-scale Percolator SVM learning is reduced to nearly only a fifth of the original runtime. Importantly, these speedups only affect the speed at which Percolator converges to a global solution and do not alter recalibration performance. The upgraded versions of both l-SVM-MFN and TRON are optimized within the Percolator codebase for multithreaded and single-thread use and are available under Apache license at bitbucket.org/jthalloran/percolator_upgrade .

Identification of risk genes related to myocardial infarction and the construction of early SVM diagnostic model

“…We screened out significant risk genes through feature selection and optimization. An SVM model was trained using ten-fold crossvalidation [18]. The SVM model is a supervised classification algorithm of machine learning.…”

Speeding Up Percolator

“…The second bottleneck is the execution time required to learn SVM parameters. Recent work 11 has tackled this bottleneck through software optimizations to Percolator's SVM learning engine, and our efforts complement and further improve upon these optimizations. On a massive data set containing over 215 million PSMs, the new version of Percolator achieves an overall speedup of 439% (81.4 h down to 18.6 h).…”

“…Finally, we optimized the CGLS solver itself using a mixture of low-level linear algebra function calls and software streamlining, as described previously. 11 Optimizations are compared against the recently described CGLS multithreaded speedup, 11 referred to as CGLS-par. In contrast to the second in our series of optimizations, which uses multiple threads to parallelize runs of CGLS at the crossvalidation level, CGLS-par instead uses multiple threads to parallelize computation within the CGLS algorithm.…”

Ion Mobility Coupled to a Time-of-Flight Mass Analyzer Combined With Fragment Intensity Predictions Improves Identification of Classical Bioactive Peptides and Small Open Reading Frame-Encoded Peptides

“…Recent advances in machine learning tools and widespread use of high throughput techniques provides a massive amount of data as a source to develop tools for every step in MSbased workflows (Bouwmeester et al, 2020). For example, the post-processing tool Percolator (Käll et al, 2007;Halloran and Rocke, 2018) integrates several features into a semi-supervised learning algorithm to improve the distinction between true and false peptide-spectrum matches. Next to that, spectrum intensity predictors, such as MS 2 PIP (Degroeve et al, 2015;Gabriels et al, 2019) and Prosit (Gessulat et al, 2019) are new models that incorporate fragment ion intensities predictions as additional features next to the standard m/z ratio during spectral library searching to increase the resolution of the identification, even in challenging workflows such as proteogenomics (Verbruggen et al, 2021).…”