Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 18 (6), 2493-2500

DO-MS: Data-Driven Optimization of Mass Spectrometry Methods

Affiliations

DO-MS: Data-Driven Optimization of Mass Spectrometry Methods

R Gray Huffman et al. J Proteome Res.

Abstract

The performance of ultrasensitive liquid chromatography and tandem mass spectrometry (LC-MS/MS) methods, such as single-cell proteomics by mass spectrometry (SCoPE-MS), depends on multiple interdependent parameters. This interdependence makes it challenging to specifically pinpoint the sources of problems in the LC-MS/MS methods and approaches for resolving them. For example, a low signal at the MS2 level can be due to poor LC separation, ionization, apex targeting, ion transfer, or ion detection. We sought to specifically diagnose such problems by interactively visualizing data from all levels of bottom-up LC-MS/MS analysis. Many software packages, such as MaxQuant, already provide such data, and we developed an open source platform for their interactive visualization and analysis: Data-driven Optimization of MS (DO-MS). We found that in many cases DO-MS not only specifically diagnosed LC-MS/MS problems but also enabled us to rationally optimize them. For example, by using DO-MS to optimize the sampling of the elution peak apexes, we increased ion accumulation times and apex sampling, which resulted in a 370% more efficient delivery of ions for MS2 analysis. DO-MS is easy to install and use, and its GUI allows for interactive data subsetting and high-quality figure generation. The modular design of DO-MS facilitates customization and expansion. DO-MS v1.0.8 is available for download from GitHub: https://github.com/SlavovLab/DO-MS . Additional documentation is available at https://do-ms.slavovlab.net .

Keywords: MaxQuant; R; Shiny; method development; optimizing mass spectrometry; quality control; single-cell analysis; single-cell proteomics by mass spectrometry; ultrasensitive proteomics; visualization.

Conflict of interest statement

The authors declare no competing financial interest.

Figures

Figure 1.
Figure 1.
DO-MS-assisted optimization of instrument parameters leads to increased apex targeting, ion delivery, and peptide identification rates. (a) Distributions of apex offsets for three runs on 90 min gradients. Each injection was of 1 μL from the same vial and corresponds to 1 × M dilution of a SCoPE-MS master sample as described by Specht et al.27 All LC parameters and instrument parameters were set to be the same except for the max fill time. (b) The relative efficiency of delivering ions for MS2 analysis was estimated by the intensities of RI. For each peptide identified across all three experiments, the RI intensity was divided by the corresponding RI intensity for 250 ms fill time, and the results for all peptides are shown as distributions on a log2 scale. (c) The number of peptides identified at each PEP threshold is shown as a rank sorted list for the three fill times. This display shows the number of peptides for all levels of confidence of identification, as quantified by the PEP. The plots from panels (a–c) can be found in the “Ion Sampling” tab and in the “Peptide Identifications” tabs of DO-MS, respectively. The plot for panel (b) was normalized to the 250 ms experiment specifically for this figure to emphasize the increased ion accumulation.
Figure 2.
Figure 2.
Diagnosing reduced peptide identification due to co-eluting contaminants. Plotting the cumulative intensities for all +1 ions detected during the survey scans (a) alongside the number of peptides identified across the gradient (b) can reveal correlations between co-eluting contaminants and reduced peptide identification. (c) Number of all detected ions by charge states. (d) Peptides were rank sorted by their PEPs to display the number of identified peptides across all levels of confidence. The plots from panels (a–c) can be found within the “Contamination” tab of DO-MS, while the plot shown in panel (d) can be found in the “Peptide Identifications” tab.
Figure 3.
Figure 3.
Controlled comparison of peptide abundances across experiments. Without controlling for the composition of the two populations being compared, trends in the data can be misread. In this case, when comparing the distribution of precursor intensities for all peptides identified in each sample (a), sample 1 appears to have more highly abundant peptides. However, when ensuring that the comparison is only based on those peptides identified in each sample (b), the opposite trend becomes apparent, namely, that the peptide species in sample 2 were more highly abundant. Both of the plots shown in Figure 3 can be found in the “Ion Sampling” tab of DO-MS.
Figure 4.
Figure 4.
Evaluating low-input samples, such as SCoPE-MS sets. (a) The distributions of rRI intensities can indicate the relative amount of peptides and the efficiency of sample preparation for each channel. (b) The matrix of pairwise correlations among all all channels of a SCoPE-MS set can be used to benchmark relative quantification within that set. In (a) we expect single-cell channels to have relative rRI intensities that are 50-fold lower than the 50-cell carrier channels (about 1.7 on log10 scale). In (b), we expect single-cell channels to correlate positively with single-cell channels and carrier channels that contain their respective cell type, while cross-cell-type correlations for single-cell channels are expected to be negative. Both of the plots shown in Figure 4 can be found in the “SCoPE-MS Diagnostics” tab of DO-MS.

Similar articles

See all similar articles

Cited by 3 PubMed Central articles

Publication types

Feedback