Label-free quantification (LFQ) proteomic data analysis from DIA-NN output files

We start with a DIA data acquisition peptide search output file the DIANN search (DIA; link to DIA-NN paper) and we trim out unused columns in the reports to simplify the analysis.

DIA-NN report restricted to:

• Protein Group
• Protein Name
• Genes
• Protein Description
• Peptide Sequence
• Sample
• Replicate
• Intensity value (counts, peptide peak area in arbitrary units)

All of the peptide intensity values (counts) are summed to the protein intensity (counts). The resulting data table is exported as:

Full_list_proteins_XXXXXXXXX-xxxxxxx.csv

Then the protein intensities (counts) of the sample replicates are averaged (mean), the standard deviation (SD), percent coefficient of variation (CV%), and Z-scores (across all samples) are calculated. The resulting data table is exported as:

Full_list_proteins_summary_XXXXXXXXX-xxxxxxx.csv

Found in the QC_files folder:

Bar plots of total proteins and peptides: The bar charts show the number of peptides or proteins identified and quantified by DIA-NN from each sample and the cumulative number for all the samples in the dataset. The protein plot also includes the number of proteins that meet the criteria for the Top3 protein quantification method.

Found in the PCA_plot folder:

PCA plot: The PCA plot shows clusters of individual sample replicates based on their similarity. The amount of explained variance contributed by the first two principal components (PC1 + PC2) is shown as the subtitle. This plot can help identify outliers and the overall precision of the data.

Scree plot: The scree plot displays the variation contributed by the top four principal components from the data.

PCA plot calculations:

1. The data is scaled with the sklearn StandardScaler fit_transform method.
2. The PCA is implemented with the sklearn PCA method. The number of principal components are limited to 4.
3. Calculate the explained variance and the cumulative variance for the top two components
4. Plot the 2D PCA graph
5. Plot the Scree graph

Coefficient of variation (CV) violin plot:

We use the Top3 quantification method (references below) to calculate the absolute protein abundance as fractions of total protein mass in each sample. Briefly, the Top3 quantification method is based on the “best flyer” hypothesis, which assumes that the specific MS signal intensity of the most intense tryptic peptides per protein is approximately constant throughout a whole proteome (ref: Ludwig et al. Mol. Cell. Proteomics 2012).

Our Top3 quantification analysis consists of:

1. Filter the DIA-NN peptide report data (from step 1) to only proteins that have three or more peptides identified across all samples
2. For each protein, rank the top 3 peptides by intensity (counts) in each of the samples
3. Calculate the mean rankings of the peptides in each protein across all samples
4. Filter the data to the three highest ranked peptides in each protein
5. Calculate protein intensity (counts) by averaging the intensity (counts) of the Top3 peptides
6. Calculate the percent of the total protein abundance ((intensity of individual protein / sum of all protein intensities in a given sample) * 100)

The resulting data tables are exported as:

Top3 full peptide list:

Top3_Full_list_peptides_used_for_quant_XXXXXXXX-xxxxxx.csv

Top3 full protein list for each replicate:

Top3_Full_list_proteins_XXXXXXXX-xxxxxx.csv

Top3 full list of proteins averaged across replicates:

Top3_Full_list_proteins_summary_XXXXXXXX-xxxxxx.csv

References:

Ludwig et al. DOI 10.1074/mcp.M111.013987

Silva et al. DOI 10.1074/mcp.M500230-MCP200

Ahrne et al. DOI 10.1002/pmic.201300135

Grossman et al. DOI 10.1016/j.jprot.2010.05.011

If a list of selected proteins is provided then a histogram for each sample is generated in .png, format with the categories of selected proteins highlighted with the background proteome:

• X-axis: log10 (% protein abundance) bins

• Y-axis: count of proteins in the bins

• Files generated: SampleID-histogram_XXXXXXXX-xxxxxx.png

• Files generated: Top3_allsamples_jitterplot_XXXXXXXX-xxxxxx.png

  Top3_allsamples_jitterplot_XXXXXXXX-xxxxxx.plotly

If a list of selected proteins is provided bar charts are generated in .png, .svg, and .plotly formats:

• X-axis: Proteins

• Y-axis: % protein abundance (from Top3 quantification method) averaged over replicates

• Error bars: standard deviation of % protein abundance (from Top3 quantification method) from the replicates

• Files generated: Full_and_select_proteins_summary_XXXXXXXX-xxxxxx.xlsx

    selectproteincategory-bar_XXXXXXXX-xxxxxx.png
  
    selectproteincategory-bar_XXXXXXXX-xxxxxx.svg
  
    selectproteincategory-bar_XXXXXXXX-xxxxxx.plotly
  

If other commonly analyzed proteins (e.g., insoluble protein diagnsotic marker proteins, proteases, heat shock proteins) are detected and quantified then a bar chart is generated in .png, .svg, and .plotly formats with only the corresponding data:

Example filenames:

insol-marker-bar_XXXXXXXX-xxxxxx.png

insol-marker-bar_XXXXXXXX-xxxxxx.svg

insol-marker-bar_XXXXXXXX-xxxxxx.plotly

If a list of selected proteins is provided strip charts are generated in .png, .svg, and .plotly formats to show the individual data points for each sample:

• X-axis: Proteins

• Y-axis: Protein intensity (counts) calculated from the mean of the top 3 peptides for each sample replicate

• Error bars: none

• Files generated: selectproteincategory-strip_XXXXXXXX-xxxxxx.png

    selectproteincategory-strip_XXXXXXXX-xxxxxx.svg
  
    selectproteincategory-strip_XXXXXXXX-xxxxxx.plotly
  

If a list of selected proteins is provided AND the sample names contain timepoint information (e.g., CJP1234_24hr-R1) line charts are generated in .png, .svg, and .plotly formats to show the individual data points for each sample:

• Sub-plot: Protein

• X-axis: Time

• Y-axis: % protein abundance (Top3 method)

• Error bars: standard deviation of % protein abundance (Top3 method) from the replicates

• Files generated: selectproteincategory-line_XXXXXXXX-xxxxxx.png

    selectproteincategory-line_XXXXXXXX-xxxxxx.svg
  
    selectproteincategory-line_XXXXXXXX-xxxxxx.plotly
  

If applicable, two samples (A and B) are selected for comparison then a Welch's t-Test is performed by using the ttest_ind_from_stats function from scipy (details here). This is comparable to the Excel function t-Test: Two-Sample Assuming Unequal Variances.

For this analysis:

• Missing values and zero abundance values are imputed with the lowest of detected (LOD) value in each sample.
• Abundance values are log2 transformed prior to the t-Test
• The False Discovery Rate (FDR; adjusted p-value; q-value) is calculated by the Benjamini-Hochberg method by using the statsmodels.stats.multitest.multipletests function.

Significantly changing proteins are defined as:

• a p-value (or adjusted p-value) < 0.05
• a fold change of > 2 (UP) or < 0.5 (DOWN)

The resulting data tables are exported as an Excel file (xlsx):

t-Test_SampleA_OVER_SampleB_XXXXXXXX-xxxxxx.csv

with five sheets corresponding to:

1. Full t-test output
2. p-value Significant UP changing proteins (p-value <0.05)
3. p-value Significant DOWN changing proteins (p-value <0.05)
4. adjusted p-value Significant UP changing proteins (adjusted p-value <0.05)
5. adjusted p-value Significant DOWN changing proteins (adjusted p-value <0.05)

If a list of selected proteins is provided two volcano plots are generated in .png, .svg, and .plotly formats (six total volcano plot visualization outputs) for the two sample comparisons:

• log2 (Fold change) vs. -log10(p-value) plots Volcano_plot_p-value_SampleA_OVER_SampleB_XXXXXXXX-xxxxxx.png

    Volcano_plot_p-value_SampleA_OVER_SampleB_XXXXXXXX-xxxxxx.svg
  
  
  
    Volcano_plot_p-value_SampleA_OVER_SampleB_XXXXXXXX-xxxxxx.plotly
  

• log2 (Fold change) vs. -log10(adjusted-p-value) plots Volcano_plot_adj-p-value_SampleA_OVER_SampleB_XXXXXXXX-xxxxxx.png

    Volcano_plot_adj-p-value_SampleA_OVER_SampleB_XXXXXXXX-xxxxxx.svg
  
  
  
    Volcano_plot_adj-p-value_SampleA_OVER_SampleB_XXXXXXXX-xxxxxx.plotly
  

The significance cutoffs are defined as:

• Fold Change = 0.5x and 2x (-1 and 1 on the log2 axis)
• p-value and adj-p-value = 0.05 (1.3 on the -log10 axis)