Differential Expression Analysis Quality Control

Author

Functional Genomics Center Zurich

Published

July 29, 2026

Samples12

Groups3

Proteins100

Visual abstract: quantified proteins are checked for missingness, variance, abundance distributions, and sample structure before differential-expression quality is assessed.

This report checks whether quantified protein abundances are suitable for differential-expression interpretation. It examines missing values, within-group variability, abundance distributions, and sample structure before the differential-expression diagnostics are read.

Input: quantified protein abundances from 12 samples, analysed with DIANN.

Missing values can reveal potential biases or technical problems in the data. Figure 1 summarizes missing protein abundance estimates per group. Panel A shows how many proteins have $0-N$ missing values. Ideally, most proteins are quantified in all samples within a group. Panel B shows the distribution of mean protein intensity by the number of missing values. Proteins without missing values usually have higher average abundance than proteins with one or more missing values because low-abundance proteins are more likely to remain undetected. Strongly overlapping distributions can point to other sources of missingness, such as large sample heterogeneity or technical problems.

Figure 1: Missing protein abundance estimates by group. Panel A: number of proteins with n missing values (nrNA). Panel B: distribution of mean within-group protein intensity by the number of missing values.

Panel A in Figure 2 shows the coefficients of variation (CVs) for all proteins computed from non-normalized data. Ideally, the within-group CV should be smaller than the CV across all samples. Panel B shows the standard deviation distribution for log2-transformed data, while panel C shows the corresponding distribution after normalization. Normalization should reduce within-group variance compared with the overall variance. If normalization increases within-group variance relative to the overall variance, the selected normalization method may not be compatible with the data.

Figure 2: Protein-level variance distributions. Panel A: coefficients of variation (CVs) within groups and across the full experiment (all) for non-normalized data. Panel B: standard deviations (SDs) for log2-transformed data. Panel C: SDs after normalization. Black dots indicate medians.

Table 1 shows the median CV and SD values for all groups and across all samples (all).

Table 1: Median (50th-percentile) coefficient of variation (CV, from non-normalized data) and standard deviation (SD) for log2-transformed (sd_log2) and normalized (sd) data, per group and across all samples (all).

what	A	B	Ctrl	All
CV	3.47	3.37	3.64	9.32
sd_log2	0.05	0.05	0.05	0.14
sd	0.06	0.05	0.06	0.14

Most proteins in a dataset are usually not differentially expressed; therefore, differences between the two groups should be centered close to zero. Figure 3 shows the distribution of group differences for all proteins. Ideally, the median of this distribution (red line) should be close to zero (green line). If the median and mode of the difference distribution are non-zero, this should be considered when interpreting the differential expression results.

Panel B in Figure 3 shows the distribution of p-values for all proteins. If the null hypothesis is true, p-values should be approximately uniformly distributed. A subset of differentially expressed proteins produces a higher frequency of small p-values. A higher frequency of large p-values close to 1 can indicate that the linear model does not describe the variance of the data well, for example because of outliers or an unmodeled source of variability.

Figure 3: Differential expression diagnostic distributions. Panel A: differences between groups for all proteins. The red dotted line marks the median fold change; the green line marks the expected median fold change. Panel B: histogram of p-values for all proteins.

The MA plot in Figure 4 helps identify whether large fold changes are concentrated among high- or low-abundance proteins. Panel A shows the group difference (y-axis) as a function of average protein abundance (x-axis). The observed fold change should not depend on protein abundance. Panel B shows the same group difference against the rank of average protein abundance.

Figure 4: MA plot diagnostics. Panel A: group difference versus average protein abundance. Panel B: group difference versus the rank of average protein abundance.

Table 2: Compact report provenance, including the source input-data reference.

Field	Value
Workunit ID	23000
Order ID	6200
Project ID	3000
Project name	n/a
Creator	runner
Created at	2026-07-29 11:12:58 UTC
Input data	https://fgcz-bfabric.uzh.ch/bfabric/
Quantification software	DIANN
Model	lm_impute
prolfquapp version	2.6.1

R version 4.6.1 (2026-06-24)
Platform: x86_64-pc-linux-gnu
Running under: Ubuntu 24.04.4 LTS

Matrix products: default
BLAS:   /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3 
LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/libopenblasp-r0.3.26.so;  LAPACK version 3.12.0

locale:
 [1] LC_CTYPE=C.UTF-8       LC_NUMERIC=C           LC_TIME=C.UTF-8       
 [4] LC_COLLATE=C.UTF-8     LC_MONETARY=C.UTF-8    LC_MESSAGES=C.UTF-8   
 [7] LC_PAPER=C.UTF-8       LC_NAME=C              LC_ADDRESS=C          
[10] LC_TELEPHONE=C         LC_MEASUREMENT=C.UTF-8 LC_IDENTIFICATION=C   

time zone: UTC
tzcode source: system (glibc)

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
  [1] RColorBrewer_1.1-3          jsonlite_2.0.0             
  [3] shape_1.4.6.1               magrittr_2.0.5             
  [5] jomo_2.7-6                  farver_2.1.2               
  [7] logistf_1.26.1              nloptr_2.2.1               
  [9] rmarkdown_2.31              GlobalOptions_0.1.4        
 [11] vctrs_0.7.3                 minqa_1.2.8                
 [13] progress_1.2.3              htmltools_0.5.9            
 [15] S4Arrays_1.12.0             forcats_1.0.1              
 [17] broom_1.0.13                cellranger_1.1.0           
 [19] SparseArray_1.12.2          mitml_0.4-5                
 [21] htmlwidgets_1.6.4           plyr_1.8.9                 
 [23] plotly_4.12.1               mime_0.13                  
 [25] lifecycle_1.0.5             iterators_1.0.14           
 [27] pkgconfig_2.0.3             Matrix_1.7-5               
 [29] R6_2.6.1                    fastmap_1.2.0              
 [31] rbibutils_2.4.1             MatrixGenerics_1.24.0      
 [33] clue_0.3-68                 digest_0.6.39              
 [35] dtplyr_1.3.3                colorspace_2.1-3           
 [37] lobstr_1.2.1                S4Vectors_0.50.1           
 [39] crosstalk_1.2.2             GenomicRanges_1.64.0       
 [41] labeling_0.4.3              httr_1.4.8                 
 [43] abind_1.4-8                 mgcv_1.9-4                 
 [45] compiler_4.6.1              withr_3.0.3                
 [47] bit64_4.8.2                 doParallel_1.0.17          
 [49] pander_0.6.6                S7_0.2.2                   
 [51] backports_1.5.1             logger_0.4.2               
 [53] UpSetR_1.4.1                prolfquasaint_0.1.5        
 [55] pan_2.0                     MASS_7.3-65                
 [57] DelayedArray_0.38.2         rjson_0.2.23               
 [59] optparse_1.8.2              tools_4.6.1                
 [61] otel_0.2.0                  nnet_7.3-20                
 [63] glue_1.8.1                  nlme_3.1-169               
 [65] grid_4.6.1                  cluster_2.1.8.2            
 [67] generics_0.1.4              operator.tools_1.6.3.1     
 [69] gtable_0.3.6                tzdb_0.5.0                 
 [71] formula.tools_1.7.1         preprocessCore_1.74.0      
 [73] tidyr_1.3.2                 data.table_1.18.4          
 [75] hms_1.1.4                   XVector_0.52.0             
 [77] BiocGenerics_0.58.1         ggrepel_0.9.8              
 [79] foreach_1.5.2               pillar_1.11.1              
 [81] stringr_1.6.0               limma_3.68.4               
 [83] circlize_0.4.18             splines_4.6.1              
 [85] dplyr_1.2.1                 lattice_0.22-9             
 [87] survival_3.8-6              bit_4.6.0                  
 [89] tidyselect_1.2.1            ComplexHeatmap_2.28.0      
 [91] knitr_1.51                  reformulas_0.4.4           
 [93] gridExtra_2.3.1             prolfquapp_2.6.1           
 [95] bookdown_0.47               IRanges_2.46.0             
 [97] Seqinfo_1.2.0               SummarizedExperiment_1.42.0
 [99] stats4_4.6.1                xfun_0.60                  
[101] prolfqua_1.7.0              Biobase_2.72.0             
[103] statmod_1.5.2               matrixStats_1.5.0          
[105] stringi_1.8.7               yaml_2.3.12                
[107] boot_1.3-32                 evaluate_1.0.5             
[109] codetools_0.2-20            tibble_3.3.1               
[111] BiocManager_1.30.27         cli_3.6.6                  
[113] affyio_1.82.0               rpart_4.1.27               
[115] arrow_25.0.0                Rdpack_2.6.6               
[117] Rcpp_1.1.2                  readxl_1.5.0               
[119] png_0.1-9                   parallel_4.6.1             
[121] ggplot2_4.0.3               readr_2.2.0                
[123] assertthat_0.2.1            prettyunits_1.2.0          
[125] lme4_2.0-6                  glmnet_5.0                 
[127] viridisLite_0.4.3           scales_1.4.0               
[129] affy_1.90.0                 purrr_1.2.2                
[131] crayon_1.5.3                writexl_1.5.4              
[133] GetoptLong_1.1.1            rlang_1.3.0                
[135] vsn_3.80.0                  mice_3.19.0

--- title: "Differential Expression Analysis Quality Control" author: "Functional Genomics Center Zurich" date: today format: html: toc-location: left grid: sidebar-width: 300px include-after-body: fgcz-plot-finder.html toc: true params: deanalyse_file: null vignette: > %\VignetteIndexEntry{Differential Expression QC} %\VignetteEngine{quarto::format} %\VignetteEncoding{UTF-8} --- ```{r setup} #| include: false knitr::opts_chunk$set( echo = FALSE, message = FALSE, warning = FALSE, out.width = "600px", fig.align = "center" ) deanalyse_file <- params$deanalyse_file if (is.null(deanalyse_file) || !nzchar(deanalyse_file)) { deanalyse_file <- file.path(tempdir(), "prolfquapp-example-deanalyse.rds") if (!file.exists(deanalyse_file)) { saveRDS(prolfquapp::example_deanalyse(), deanalyse_file) } } dea <- readRDS(deanalyse_file) cfg <- dea$prolfq_app_config report_html_attr <- function(x) { value <- prolfquapp:::.report_provenance_scalar(x) value <- gsub("&", "&", value, fixed = TRUE) value <- gsub('"', """, value, fixed = TRUE) value <- gsub("<", "<", value, fixed = TRUE) value <- gsub(">", ">", value, fixed = TRUE) value } project_spec <- cfg$project_spec report_metadata <- prolfquapp:::.report_provenance( project_spec = project_spec, software = cfg$software, model = dea$default_model ) ``` ```{r report-metadata-marker} #| results: asis cat(sprintf( paste0( '<div id="fgcz-report-metadata" hidden ', 'data-project-id="%s" data-order-id="%s" data-workunit-id="%s" ', 'data-generated-by="%s" data-generated-at="%s"></div>\n' ), report_html_attr(report_metadata$project_id), report_html_attr(report_metadata$order_id), report_html_attr(report_metadata$workunit_id), report_html_attr(report_metadata$creator), report_html_attr(report_metadata$created_at) )) ``` ::: {.panel-tabset} # Overview ```{r overview-data-summary} #| results: asis cat(prolfquapp:::.report_overview_cards(dea$lfq_data_raw)) ``` <img src="`r knitr::image_uri(system.file("report-assets", "differential-expression-qc.png", package = "prolfquapp"))`" alt="Visual abstract: quantified proteins are checked for missingness, variance, abundance distributions, and sample structure before differential-expression quality is assessed." style="width: 100%; max-height: 22rem; object-fit: contain;"> This report checks whether quantified protein abundances are suitable for differential-expression interpretation. It examines missing values, within-group variability, abundance distributions, and sample structure before the differential-expression diagnostics are read. Input: quantified protein abundances from `r nrow(dea$lfq_data_raw$factors())` samples, analysed with `r report_metadata$software`. # Missing Values Missing values can reveal potential biases or technical problems in the data. @fig-missing-protein summarizes missing protein abundance estimates per group. Panel A shows how many proteins have $0-N$ missing values. Ideally, most proteins are quantified in all samples within a group. Panel B shows the distribution of mean protein intensity by the number of missing values. Proteins without missing values usually have higher average abundance than proteins with one or more missing values because low-abundance proteins are more likely to remain undetected. Strongly overlapping distributions can point to other sources of missingness, such as large sample heterogeneity or technical problems. ```{r} #| label: fig-missing-protein #| fig-cap: "Missing protein abundance estimates by group. Panel A: number of proteins with n missing values (nrNA). Panel B: distribution of mean within-group protein intensity by the number of missing values." #| fig-height: 5 #| fig-width: 8 #| out-width: "600px" pl <- dea$lfq_data_raw$get_Plotter() p2 <- pl$missigness_histogram() sr <- dea$lfq_data_raw$get_Summariser() p1 <- sr$plot_missingness_per_group() gridExtra::grid.arrange( p1 + ggplot2::labs(tag = "A"), p2 + ggplot2::labs(tag = "B"), nrow = 2 ) ``` # Variance Panel A in @fig-sd-violin shows the coefficients of variation (CVs) for all proteins computed from non-normalized data. Ideally, the within-group CV should be smaller than the CV across all samples. Panel B shows the standard deviation distribution for log2-transformed data, while panel C shows the corresponding distribution after normalization. Normalization should reduce within-group variance compared with the overall variance. If normalization increases within-group variance relative to the overall variance, the selected normalization method may not be compatible with the data. ```{r} #| label: fig-sd-violin #| fig-cap: "Protein-level variance distributions. Panel A: coefficients of variation (CVs) within groups and across the full experiment (all) for non-normalized data. Panel B: standard deviations (SDs) for log2-transformed data. Panel C: SDs after normalization. Black dots indicate medians." #| fig-height: 3 #| fig-width: 7 #| out-width: "600px" stR <- dea$lfq_data_raw$get_Stats() pA <- stR$violin() + ggplot2::theme_classic() + ggplot2::labs(tag = "A") + ggplot2::theme(axis.text.x = ggplot2::element_text(angle = 45, hjust = 1)) st <- dea$lfq_data$get_Stats() pC <- st$violin() + ggplot2::theme_classic() + ggplot2::labs(tag = "C") + ggplot2::theme(axis.text.x = ggplot2::element_text(angle = 45, hjust = 1)) lfqData <- dea$lfq_data_raw tr <- lfqData$get_Transformer() tmp <- tr$log2()$lfq stback <- tmp$get_Stats() pB <- stback$violin() + ggplot2::theme_classic() + ggplot2::labs(tag = "B") + ggplot2::theme(axis.text.x = ggplot2::element_text(angle = 45, hjust = 1)) gridExtra::grid.arrange(pA, pB, pC, nrow = 1) ``` @tbl-cv shows the median CV and SD values for all groups and across all samples (`all`). ```{r} #| label: tbl-cv #| tbl-cap: "Median (50th-percentile) coefficient of variation (CV, from non-normalized data) and standard deviation (SD) for log2-transformed (sd_log2) and normalized (sd) data, per group and across all samples (all)." resR <- stR$stats_quantiles() res <- st$stats_quantiles() reslog2 <- stback$stats_quantiles() C <- dplyr::bind_rows( CV = resR$wide |> dplyr::filter(probs == 0.5) |> round(digits = 2), sd_log2 = reslog2$wide |> dplyr::filter(probs == 0.5) |> round(digits = 2), sd = res$wide |> dplyr::filter(probs == 0.5) |> round(digits = 2) ) C <- C |> dplyr::mutate(what = c("CV", "sd_log2", "sd"), .before = 1) C$probs <- NULL knitr::kable(C) ``` # Differential Expression Most proteins in a dataset are usually not differentially expressed; therefore, differences between the two groups should be centered close to zero. @fig-fc-density shows the distribution of group differences for all proteins. Ideally, the median of this distribution (red line) should be close to zero (green line). If the median and mode of the difference distribution are non-zero, this should be considered when interpreting the differential expression results. Panel B in @fig-fc-density shows the distribution of p-values for all proteins. If the null hypothesis is true, p-values should be approximately uniformly distributed. A subset of differentially expressed proteins produces a higher frequency of small p-values. A higher frequency of large p-values close to 1 can indicate that the linear model does not describe the variance of the data well, for example because of outliers or an unmodeled source of variability. ```{r} #| label: fig-fc-density #| fig-cap: "Differential expression diagnostic distributions. Panel A: differences between groups for all proteins. The red dotted line marks the median fold change; the green line marks the expected median fold change. Panel B: histogram of p-values for all proteins." #| fig-width: 8 #| fig-height: 4 #| out-width: "600px" cpl <- dea$contrast_results[[dea$default_model]]$get_Plotter( fc_threshold = cfg$processing_options$diff_threshold, fdr_threshold = cfg$processing_options$FDR_threshold) p1 <- cpl$histogram_diff() + ggplot2::labs(tag = "A") p2 <- cpl$histogram()$p.value + ggplot2::labs(tag = "B") gridExtra::grid.arrange(p1, p2) ``` The MA plot in @fig-ma helps identify whether large fold changes are concentrated among high- or low-abundance proteins. Panel A shows the group difference (y-axis) as a function of average protein abundance (x-axis). The observed fold change should not depend on protein abundance. Panel B shows the same group difference against the rank of average protein abundance. ```{r} #| label: fig-ma #| fig-cap: "MA plot diagnostics. Panel A: group difference versus average protein abundance. Panel B: group difference versus the rank of average protein abundance." #| fig-width: 8 #| fig-height: 5 #| out-width: "600px" cpl <- dea$contrast_results[[dea$default_model]]$get_Plotter( fc_threshold = cfg$processing_options$diff_threshold, fdr_threshold = cfg$processing_options$FDR_threshold) ppb <- cpl$ma_plot(rank = FALSE) + ggplot2::labs(tag = "A") ppc <- cpl$ma_plot(fc = cfg$processing_options$diff_threshold, rank = TRUE) + ggplot2::labs(tag = "B") gridExtra::grid.arrange(ppb, ppc, ncol = 1) ``` # Session Info ::: {.panel-tabset} ## Report provenance ```{r} #| label: tbl-report-metadata #| tbl-cap: "Compact report provenance, including the source input-data reference." knitr::kable(prolfquapp:::.report_provenance_table(report_metadata)) ``` ## R session info ```{r session-info} sessionInfo() ``` ::: :::

Differential Expression Analysis Quality Control

Download

Find figures & tables