High sensitivity dia-PASEF proteomics with DIA-NN and FragPipe

Abstract

The dia-PASEF technology exploits ion mobility separation for high-sensitivity analysis of complex proteomes. Here, we demonstrate neural network-based processing of the ion mobility data, which we implement in the DIA-NN software suite. Using spectral libraries generated with the MSFragger-based FragPipe computational platform, the DIA-NN analysis of dia-PASEF raw data increases the proteomic depth by up to 69% compared to the originally published dia-PASEF workflow. For example, we quantify over 5200 proteins from 10ng of HeLa peptides separated with a 95-minute nanoflow gradient, and over 5000 proteins from 200ng using a 4.8-minute separation with an Evosep One system. In complex samples, featuring a mix of human and yeast lysates, the workflow detects over 11700 proteins in single runs acquired with a 100-minute nanoflow gradient, while demonstrating quantitative precision. Hence, the combination of FragPipe and DIA-NN provides a simple-to-use software platform for dia-PASEF data analysis, yielding significant gains in high-sensitivity proteomics.

Comparison of TIMS-PASEF quantitative proteomics data-analysis workflows using FragPipe, DIA-NN, and Spectronaut from a user’s perspective

“…The highest data completeness (98.8% and 98.5%) was achieved by the FragPipe+DIA-NN workflows. These results improve on those reported by Demichev et al [8], where data completeness ranged from 94% to 98%. On the other hand, LFQ with IonQuant using 2 min ions, resulted in the highest percentage of missing values (8.4%).…”

“…Median CVs were below 9% in all workflows, with Spectronaut in the range 7.9 to 8.8%, and label free DIA-NN with 7.2%. Library-based DIA-NN median CVs obtained (6.5% and 6%) are in the range of those reported by Demichev et al [8] (2.7% to 9.5%). FragPipe's IonQuant performed the best in terms of median CV (3.6% and 4.3% for the 2 and 1 minimum ion analysis, respectively).…”

“…Regarding diaPASEF data, Demichev et al [8] reported the good performance of the combination of building the library in MSFragger and then using this library to analyze the diaPASEF runs in DIA-NN. However, a study comparing the different analysis possibilities offered by the latest software developments optimized for diaPASEF data analysis is still missing.…”

Trapped Ion Mobility Spectrometry and Parallel Accumulation–Serial Fragmentation in Proteomics

“…The more recently developed DIA-NN software 90 uses neural networks to match the MS/MS data to a library and has recently been extended for dia-PASEF. Reanalysis of our original dia-PASEF data with DIA-NN and a library generated by MSFragger significantly increased protein identifications while improving quantification, including the identification of more than 5,000 proteins in 5 min EvoSep runs 91 . With MaxDIA, dia-PASEF data can now also be analyzed within the MaxQuant environment 92 .…”

In-Depth Benchmarking of DIA-type Proteomics Data Analysis Strategies Using a Large-Scale Benchmark Dataset Comprising Inter-Patient Heterogeneity

“…For the sake of simplicity we focused our statistical analyses on the comparison between the two lowest E. coli spike-in conditions (Figure 1). This also represents the greatest challenge to any DIA analysis software as quantitations are usually less precise for lowly abundant proteins 16,17 .…”