DIA-NN: neural networks and interference correction enable deep proteome coverage in high throughput

Abstract

We present an easy-to-use integrated software suite, DIA-NN, that exploits deep neural networks and new quantification and signal correction strategies for the processing of data-independent acquisition (DIA) proteomics experiments. DIA-NN improves the identification and quantification performance in conventional DIA proteomic applications, and is particularly beneficial for highthroughput applications, as it is fast and enables deep and confident proteome coverage when employed in combination with fast chromatographic methods. Proteomics provides the functional links between the genome and metabolome of a cell, and is rapidly gaining importance within both personalised medicine and the emerging field of data-driven biology 1-4. The generation of data is hampered, however, by the inherent complexity of the proteome. In mass spectrometry-based (bottom-up) proteomics, this complexity leads to stochasticity in peptide detection, reducing the proteomic sampling depth 5,6. A popular solution to these issues is to decrease sample complexity by proteome or peptidome pre-fractionation. Extensive fractionation promotes excellent proteome coverage, Users may view, print, copy, and download text and data-mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use:

Clinical classifiers of COVID-19 infection from novel ultra-high-throughput proteomics

“…Firstly, on this type of fast chromatography, conventional mass spectrometric acquisition schemes do not reach sufficient sampling velocity in data-dependent mode (peaks elute too fast). Secondly, when using data-independent acquisition schemes which do not sample each peak individually, conventional software cannot deconvolute the interference-rich short-gradient data produced (Demichev et al, 2020;Messner et al, 2019) . We were able to overcome these issues, and present an acquisition scheme that bases on 5-minute water to acetonitrile chromatographic gradients at a flow rate of 800 μl/min.…”

“…The high-flow setup used achieved a median peak Full Width at Half Maximum (FWHM) of 3 seconds with a 20-minute gradient length. For comparison, an extensively optimized micro-flow LC (Demichev et al, 2020;Messner et al, 2019) , achieved a FWHM of 5 seconds, at the same gradient length ( Figure 2c). Thus, high-flow gradients as fast as 5 minutes resulted in peak capacities comparable to the highly optimized 20-minute micro-flow setup (Demichev et al, 2020;Messner et al, 2019) (Figure 2d).…”

Parallel factor analysis enables quantification and identification of highly-convolved data independent-acquired protein spectra

“…As all false identities were based on fragments that did not originate from the nonrandomized spectral library, considering the false positive rate underestimation of the conventional tools, up to 70% (421 of the 601 common proteins shown in Figure 2d) of proteins identified at 1% FDR with tools other than PARADIAS are thus put into the question. The quantification based on a PARADIAS library, built using conventional methods Demichev et al , 2019) with an FDR threshold of 1%, resulted in precise ratios between the two mixtures A and B in the dataset (Figure 2e), showing precise quantification using a spectral library constructed directly from the decomposed spectra. DIA spectra are still proteomic "dark matter"…”

“…Despite this, an inherent issue is related to the exhaustive fragmentation of the specific mass range using defined isolation windows or "swaths" (Gillet et al , 2012) . Due to the width of these windows, fragment signals are highly overlapped or "convolved", with multiple precursors falling in the same window, producing a set of highly overlapping ion mass spectra (Pappireddi et al , 2019;Peckner et al , 2018;Demichev et al , 2019) . A computational solution to deconvolve such data would expand the coverage and efficacy of the DIA approach.…”

“…The current standard approach for DIA analysis is targeted quantification of the acquired fragment data using spectral libraries containing fragmentation information for a particular peptide (Bruderer et al , 2015;Demichev et al , 2019;Röst et al , 2014;Peckner et al , 2018) . Library generation is however time-consuming, specific to the instrument, chromatography, and experimental condition, ideally requiring physical sample fractionation complemented with shotgun spectra acquisition (Schubert et al , 2015) .…”