Interrogating data-independent acquisition LC-MS/MS for affinity proteomics

Abstract

Data-Independent Acquisition (DIA) LC-MS/MS is an attractive partner for co-immunoprecipitation (co-IP) and affinity proteomics in general. Reducing the variability of quantitation by DIA could increase the statistical contrast for detecting specific interactors versus what has been achieved in Data-Dependent Acquisition (DDA). By interrogating affinity proteomes featuring both DDA and DIA experiments, we sought to evaluate the spectral libraries, the missingness of protein quantity tables, and the CV of protein quantities in six studies representing three different instrument manufacturers. We examined four contemporary bioinformatics workflows for DIA: FragPipe, DIA-NN, Spectronaut, and MaxQuant. We determined that (1) identifying spectral libraries directly from DIA experiments works well enough that separate DDA experiments do not produce larger spectral libraries when given equivalent instrument time; (2) experiments involving mock pull-downs or IgG controls may feature such indistinct signals that contemporary software will struggle to quantify them; (3) measured CV values were well controlled by Spectronaut and DIA-NN (and FragPipe, which implements DIA-NN for the quantitation step); and (4) when FragPipe builds spectral libraries and quantifies proteins from DIA experiments rather than performing both operations in DDA experiments, the DIA route results in a larger number of proteins quantified without missing values as well as lower CV for measured protein quantities.

Supplementary information: The online version contains supplementary material available at 10.1007/s42485-024-00166-4.

Keywords: Affinity enrichment; Bioinformatics; Co-immunoprecipitation; Data-independent acquisition; Label-free quantitation.

Fig. 1

Four different DIA identification and quantitation workflows were evaluated: DIA-NN, FragPipe, Spectronaut, and MaxQuant (from left to right). Each produced a spectral library in a different format (green cylinders). Each reported protein quantity matrices to text tables (violet folders), with FragPipe using the DIA-NN quantitation engine. Whether spectral libraries were derived from DDA or DIA experiments, these libraries were used to quantify the corresponding DIA experiments

Fig. 2

This UpSet plot reveals how the six spectral libraries derived from IP Inputs “C” overlap in the peptide sequences they represent. The bars at the lower left represent the numbers of distinct peptide sequences in each library. The dots on the lines below the main graph specify which spectral libraries contain the peptides for a particular intersection of libraries. The single dot under the second bar (“6162”) indicates that these peptides were identified only by the Spectronaut search against DIA data. The sizes of the bars in the main graph represent the number of peptide sequences in each intersection

Fig. 3

Batch 2 of the ID4 and IgG control pull-downs on the SCIEX 6600 produced spectral libraries that were far more diverse than in batch 1. For WIFFs representing the ID4 co-IP (right), the DDA experiments identified more distinct peptides than the DIA experiments. The IgG control (left), however, yielded better peptide diversity for the DIA experiments than for the DDAs

Fig. 4

The nanoparticle-enriched experiments for lysosomal proteomics illustrate the distinction among the numbers of proteins identified in the spectral library, the numbers of proteins that have at least one quantitative value reported, and the numbers of proteins that have all quantitative values reported / no missing values. All these quantitative statistics reflect the application of spectral libraries to quantify DIA experiments, whether the “Library Source” was DDA or DIA

Fig. 5

The LINE-1 data provide a direct comparison of two DIA techniques on the same samples in the same instrument: High-Resolution MS1 and “Variabele Vensters.” The proteins quantified in all samples for each cohort were separated into quintiles based on the sum of reported intensities, and these images compare the coefficients of variation for the top, middle, and bottom quintiles. The IgG samples were particularly challenging due to their low peptide diversity; surprisingly, FragPipe did not output quantities for any proteins in the IgG HRMS1 cohorts even though it could identify hundreds of peptides (552 and 669 for HEK293T and N2102Ep cell lines, respectively)