1 Purpose
2 Simulate two-factor data and encode subgroups
3 Define contrasts
4 Protein-input facades
5 Protein-level volcano comparison
- 5.1 Standard facades
- 5.2 Imputation and missingness facades
6 Looking at the strongest protein-level hits
7 Proteins that could not be estimated
- 7.1 Per-sample intensities of the missing proteins
- 7.2 Estimates for the missing proteins from lm_impute and limma_impute
8 Peptide-input facades
9 Remarks
10 Session Info

1 Purpose

This vignette demonstrates the same build_contrast_analysis() facade workflow on a dataset with two factors (e.g. treatment and genotype), yielding four subgroups: C_WT, T_WT, C_KO, T_KO. We define contrasts:

T_C_gv_WT = T_WT - C_WT
T_C_gv_KO = T_KO - C_KO
WT_KO_comp = T_C_gv_WT - T_C_gv_KO

We start from simulated two-factor peptide data, derive a single subgroup factor, and run both protein-input and peptide-input facades with the same contrast specification. See ContrastFacades.Rmd for the one-factor introduction.

2 Simulate two-factor data and encode subgroups

dd <- sim_lfq_data_2factor_config(
  Nprot = 80,
  with_missing = TRUE,
  weight_missing = 1,
  PEPTIDE = TRUE,
  seed = 7
)

cfg_subgroup <- dd$config$clone(deep = TRUE)
data_subgroup <- dd$data |>
  dplyr::mutate(
    Condition = dplyr::recode(Treatment, A = "C", B = "T"),
    Genotype = dplyr::recode(Background, Z = "WT", X = "KO"),
    Subgroup = paste(Condition, Genotype, sep = "_")
  )

cfg_subgroup$factors <- c(Subgroup = "Subgroup")
cfg_subgroup$factor_depth <- 1

lfq_peptide_2f <- LFQData$new(data_subgroup, cfg_subgroup)
lfq_peptide_2f <- lfq_peptide_2f$get_Transformer()$log2()$lfq

lfq_protein_2f <- lfq_peptide_2f$get_Aggregator()$aggregate()

lfq_peptide_2f$hierarchy_keys()

## [1] "protein_Id" "peptide_Id"

lfq_protein_2f$hierarchy_keys()

## [1] "protein_Id"

lfq_protein_2f$nr_children_col()

## [1] "nr_children_protein_Id"

3 Define contrasts

contrasts_2f <- c(
  T_C_gv_WT = "SubgroupT_WT - SubgroupC_WT",
  T_C_gv_KO = "SubgroupT_KO - SubgroupC_KO",
  WT_KO_comp = "T_C_gv_WT - T_C_gv_KO"
)

Three contrasts: two within-genotype treatment effects and their difference (interaction-like comparison).

4 Protein-input facades

The following facades require aggregated input. firth is included here because it can be fitted directly on aggregated protein input.

fa_lm_2f <- build_contrast_analysis(
  lfq_protein_2f,
  "~ Subgroup",
  contrasts_2f,
  method = "lm"
)

fa_limma_2f <- build_contrast_analysis(
  lfq_protein_2f,
  "~ Subgroup",
  contrasts_2f,
  method = "limma"
)

fa_limma_impute_2f <- build_contrast_analysis(
  lfq_protein_2f,
  "~ Subgroup",
  contrasts_2f,
  method = "limma_impute"
)

fa_deqms_2f <- build_contrast_analysis(
  lfq_protein_2f,
  "~ Subgroup",
  contrasts_2f,
  method = "deqms"
)

fa_lm_impute_2f <- build_contrast_analysis(
  lfq_protein_2f,
  "~ Subgroup",
  contrasts_2f,
  method = "lm_impute"
)

fa_firth_protein_2f <- build_contrast_analysis(
  lfq_protein_2f,
  "~ Subgroup",
  contrasts_2f,
  method = "firth"
)

Because all protein-input facades share the same interface and report protein-level contrasts, their outputs can be combined directly.

# Proteins missing in the baseline lm facade (used to flag rescued proteins)
lm_missing_ids_2f <- fa_lm_2f$get_missing() |>
  dplyr::select(protein_Id, contrast) |>
  dplyr::mutate(rescued = TRUE)

results_protein_2f <- bind_rows(
  fa_lm_2f$get_contrasts(),
  fa_limma_2f$get_contrasts(),
  fa_limma_impute_2f$get_contrasts(),
  fa_lm_impute_2f$get_contrasts(),
  fa_deqms_2f$get_contrasts(),
  fa_firth_protein_2f$get_contrasts()
) |>
  dplyr::select(dplyr::any_of(c(
    "modelName", "estimate_type", "protein_Id", "contrast", "avgAbd", "diff", "FDR",
    "statistic", "std.error", "df", "p.value", "conf.low", "conf.high",
    "sigma"
  ))) |>
  dplyr::left_join(lm_missing_ids_2f, by = c("protein_Id", "contrast")) |>
  dplyr::mutate(
    rescued = dplyr::coalesce(rescued, FALSE),
    significant = FDR < 0.1 & abs(diff) > 0.5
  )

results_protein_2f |>
  dplyr::count(modelName, name = "n_results")

## # A tibble: 6 × 2
##   modelName    n_results
##   <chr>            <int>
## 1 deqms              232
## 2 firth              240
## 3 limma              232
## 4 limma_impute       240
## 5 lm                 232
## 6 lm_impute          240

The modelName column is the facade key. For facades that combine several underlying result types, such as lm_impute, the estimate_type column tells you how each row was produced (observed or lod_imputed).

results_protein_2f |>
  dplyr::count(modelName, contrast, estimate_type, name = "n_results")

## # A tibble: 24 × 4
##    modelName    contrast   estimate_type n_results
##    <chr>        <chr>      <chr>             <int>
##  1 deqms        T_C_gv_KO  observed             79
##  2 deqms        T_C_gv_WT  observed             77
##  3 deqms        WT_KO_comp observed             76
##  4 firth        T_C_gv_KO  observed             80
##  5 firth        T_C_gv_WT  observed             80
##  6 firth        WT_KO_comp observed             80
##  7 limma        T_C_gv_KO  observed             79
##  8 limma        T_C_gv_WT  observed             77
##  9 limma        WT_KO_comp observed             76
## 10 limma_impute T_C_gv_KO  lod_imputed           4
## # ℹ 14 more rows

5 Protein-level volcano comparison

5.1 Standard facades

standard_facades_2f <- c("lm", "limma", "deqms")
results_standard_2f <- results_protein_2f |>
  dplyr::filter(modelName %in% standard_facades_2f)

ggplot(results_standard_2f, aes(x = diff, y = -log10(p.value), color = significant)) +
  geom_point(alpha = 0.6, size = 1.5) +
  facet_grid(contrast ~ modelName, scales = "free_y") +
  geom_vline(xintercept = c(-0.5, 0.5), linetype = "dashed", color = "grey60") +
  geom_hline(yintercept = -log10(0.1), linetype = "dashed", color = "grey60") +
  scale_color_manual(values = c(`TRUE` = "firebrick", `FALSE` = "grey70")) +
  labs(x = "log2 fold change", y = "-log10(p.value)", color = "FDR < 0.1\nand |diff| > 0.5") +
  theme_minimal(base_size = 12)

(#fig:volcano_standard_2f)Volcano plots for the standard protein-level facades (lm, limma, deqms).

5.2 Imputation and missingness facades

Rescued proteins (missing in plain lm) are shown as large red diamonds so they stand out clearly. This group includes the LOD imputation facades (lm_impute, limma_impute) and the firth logistic regression facade which models missingness directly.

impute_facades_2f <- c("lm_impute", "limma_impute", "firth")
results_impute_2f <- results_protein_2f |>
  dplyr::filter(modelName %in% impute_facades_2f) |>
  dplyr::mutate(modelName = factor(modelName, levels = impute_facades_2f))

ggplot(results_impute_2f, aes(x = diff, y = -log10(p.value))) +
  geom_point(data = dplyr::filter(results_impute_2f, !rescued),
             aes(color = significant), alpha = 0.5, size = 1.2) +
  geom_point(data = dplyr::filter(results_impute_2f, rescued),
             color = "red", shape = 18, size = 5, alpha = 1) +
  facet_grid(contrast ~ modelName, scales = "free_y") +
  geom_vline(xintercept = c(-0.5, 0.5), linetype = "dashed", color = "grey60") +
  geom_hline(yintercept = -log10(0.1), linetype = "dashed", color = "grey60") +
  scale_color_manual(values = c(`TRUE` = "firebrick", `FALSE` = "grey70")) +
  labs(x = "log2 fold change", y = "-log10(p.value)", color = "FDR < 0.1\nand |diff| > 0.5") +
  theme_minimal(base_size = 12)

(#fig:volcano_impute_2f)Volcano plots for the imputation and missingness facades. Large red diamonds mark proteins rescued (missing in plain lm).

6 Looking at the strongest protein-level hits

results_protein_2f |>
  dplyr::group_by(modelName, contrast) |>
  dplyr::slice_min(order_by = p.value, n = 5, with_ties = FALSE) |>
  dplyr::ungroup() |>
  dplyr::select(modelName, estimate_type, contrast, protein_Id, diff, p.value, FDR)

## # A tibble: 90 × 7
##    modelName estimate_type contrast  protein_Id    diff  p.value      FDR
##    <chr>     <chr>         <chr>     <chr>        <dbl>    <dbl>    <dbl>
##  1 deqms     observed      T_C_gv_KO qc8ZRR~4628  0.628 2.16e-13 1.71e-11
##  2 deqms     observed      T_C_gv_KO H7jUye~2322  0.493 2.01e-10 7.94e- 9
##  3 deqms     observed      T_C_gv_KO zO74Fn~0389 -0.342 4.69e-10 1.01e- 8
##  4 deqms     observed      T_C_gv_KO lRuJ5o~6058  0.542 5.11e-10 1.01e- 8
##  5 deqms     observed      T_C_gv_KO uLYRH7~7195  0.588 1.51e- 9 2.35e- 8
##  6 deqms     observed      T_C_gv_WT uLYRH7~7195 -0.904 5.36e-11 4.12e- 9
##  7 deqms     observed      T_C_gv_WT qc8ZRR~4628  0.420 1.88e-10 7.24e- 9
##  8 deqms     observed      T_C_gv_WT HnBvvB~5903  0.335 1.20e- 9 3.07e- 8
##  9 deqms     observed      T_C_gv_WT IPV3OT~3010  0.467 2.42e- 9 4.66e- 8
## 10 deqms     observed      T_C_gv_WT lCgO3j~8152  0.402 4.22e- 9 6.41e- 8
## # ℹ 80 more rows

7 Proteins that could not be estimated

Every facade has a get_missing() method that returns the protein x contrast pairs present in the input data but absent from get_contrasts(). This makes it easy to see which proteins each method fails on and to compare coverage.

missing_all_2f <- dplyr::bind_rows(
  fa_lm_2f$get_missing() |> dplyr::mutate(facade = "lm"),
  fa_limma_2f$get_missing() |> dplyr::mutate(facade = "limma"),
  fa_limma_impute_2f$get_missing() |> dplyr::mutate(facade = "limma_impute"),
  fa_lm_impute_2f$get_missing() |> dplyr::mutate(facade = "lm_impute"),
  fa_deqms_2f$get_missing() |> dplyr::mutate(facade = "deqms"),
  fa_firth_protein_2f$get_missing() |> dplyr::mutate(facade = "firth")
)

missing_all_2f |>
  dplyr::count(facade, contrast, name = "n_missing") |>
  knitr::kable(caption = "Number of missing protein x contrast pairs per facade")

(#tab:missing_protein_facades_2f)Number of missing protein x contrast pairs per facade
facade	contrast	n_missing
deqms	T_C_gv_KO	1
deqms	T_C_gv_WT	3
deqms	WT_KO_comp	4
limma	T_C_gv_KO	1
limma	T_C_gv_WT	3
limma	WT_KO_comp	4
lm	T_C_gv_KO	1
lm	T_C_gv_WT	3
lm	WT_KO_comp	4

7.1 Per-sample intensities of the missing proteins

missing_proteins_2f <- unique(missing_all_2f$protein_Id)

if (length(missing_proteins_2f) > 0) {
  lfq_protein_2f$data_long() |>
    dplyr::filter(protein_Id %in% missing_proteins_2f) |>
    dplyr::select(protein_Id, sampleName,
                  !!rlang::sym(lfq_protein_2f$response())) |>
    tidyr::pivot_wider(names_from = sampleName,
                       values_from = !!rlang::sym(lfq_protein_2f$response())) |>
    knitr::kable(digits = 2, caption = "Per-sample intensities of proteins that could not be estimated")
}

(#tab:missing_intensities_2f)Per-sample intensities of proteins that could not be estimated
protein_Id	A_V1	A_V2	A_V3	A_V4	B_V1	B_V2	B_V3	B_V4	C_V1	C_V2	C_V3	C_V4	Ctrl_V1	Ctrl_V2	Ctrl_V3	Ctrl_V4
QQg7IC~3558	3.98	NA	NA	3.83	4.34	4.31	4.36	4.31	4.38	4.31	NA	4.41	NA	NA	NA	NA
hjVK4f~9433	NA	NA	3.88	NA	4.42	4.35	4.28	4.29	NA	NA	NA	NA	NA	3.68	NA	3.86
mVseto~9392	4.40	4.50	4.47	NA	4.67	4.63	4.66	4.69	4.62	4.72	4.54	4.64	NA	NA	NA	NA
zvzYsk~2881	NA	3.79	3.92	NA	NA	NA	NA	NA	NA	3.83	NA	3.92	4.18	4.17	NA	NA

The missing cells (NA) explain why these proteins cannot be estimated — they lack observations in one or more groups. The lm_impute and limma_impute facades re-fit after imputing individual values with the LOD and borrowing covariance from successful fits. These should have fewer missing proteins than plain lm or limma.

7.2 Estimates for the missing proteins from `lm_impute` and `limma_impute`

For proteins that plain lm could not estimate, the imputation-based facades can still produce contrast results. The table below shows these rescued estimates side by side.

lm_missing_proteins_2f <- fa_lm_2f$get_missing()$protein_Id |> unique()

if (length(lm_missing_proteins_2f) > 0) {
  rescued_2f <- results_protein_2f |>
    dplyr::filter(
      protein_Id %in% lm_missing_proteins_2f,
      modelName %in% c("lm_impute", "limma_impute")
    ) |>
    dplyr::arrange(protein_Id, contrast, modelName)

  rescued_2f |>
    knitr::kable(
      digits = 3,
      caption = "Contrast estimates from lm_impute and limma_impute for proteins that plain lm could not estimate"
    )
} else {
  cat("All proteins were estimable by plain lm — no rescued estimates to show.")
}

(#tab:rescued_estimates_2f)Contrast estimates from lm_impute and limma_impute for proteins that plain lm could not estimate
modelName	estimate_type	protein_Id	contrast	avgAbd	diff	FDR	statistic	std.error	df	p.value	conf.low	conf.high	sigma	rescued	significant
limma_impute	lod_imputed	QQg7IC~3558	T_C_gv_KO	4.125	0.411	0.002	5.353	0.077	8.398	0.001	0.235	0.587	0.109	FALSE	FALSE
lm_impute	lod_imputed	QQg7IC~3558	T_C_gv_KO	4.125	0.411	0.001	5.648	0.073	8.380	0.000	0.245	0.578	0.103	FALSE	FALSE
limma_impute	lod_imputed	QQg7IC~3558	T_C_gv_WT	4.075	0.351	0.005	4.568	0.077	8.398	0.002	0.175	0.526	0.109	TRUE	FALSE
lm_impute	lod_imputed	QQg7IC~3558	T_C_gv_WT	4.075	0.351	0.003	4.819	0.073	8.380	0.001	0.184	0.517	0.103	TRUE	FALSE
limma_impute	lod_imputed	QQg7IC~3558	WT_KO_comp	0.381	-0.060	0.688	-0.556	0.109	8.398	0.593	-0.309	0.188	0.109	TRUE	FALSE
lm_impute	lod_imputed	QQg7IC~3558	WT_KO_comp	0.381	-0.060	0.665	-0.586	0.103	8.380	0.573	-0.296	0.175	0.103	TRUE	FALSE
limma_impute	lod_imputed	hjVK4f~9433	T_C_gv_KO	4.118	0.435	0.003	5.705	0.076	6.398	0.001	0.251	0.620	0.108	FALSE	FALSE
lm_impute	lod_imputed	hjVK4f~9433	T_C_gv_KO	4.118	0.435	0.002	5.954	0.073	6.380	0.001	0.259	0.612	0.103	FALSE	FALSE
limma_impute	lod_imputed	hjVK4f~9433	T_C_gv_WT	3.900	0.000	1.000	0.000	0.076	6.398	1.000	-0.184	0.184	0.108	TRUE	FALSE
lm_impute	lod_imputed	hjVK4f~9433	T_C_gv_WT	3.900	0.000	1.000	0.000	0.073	6.380	1.000	-0.176	0.176	0.103	TRUE	FALSE
limma_impute	lod_imputed	hjVK4f~9433	WT_KO_comp	0.218	-0.435	0.015	-4.034	0.108	6.398	0.006	-0.696	-0.175	0.108	TRUE	FALSE
lm_impute	lod_imputed	hjVK4f~9433	WT_KO_comp	0.218	-0.435	0.013	-4.210	0.103	6.380	0.005	-0.685	-0.186	0.103	TRUE	FALSE
limma_impute	lod_imputed	mVseto~9392	T_C_gv_KO	4.519	0.280	0.002	4.803	0.058	10.398	0.001	0.151	0.409	0.109	FALSE	FALSE
lm_impute	lod_imputed	mVseto~9392	T_C_gv_KO	4.519	0.280	0.001	5.101	0.055	10.380	0.000	0.158	0.402	0.103	FALSE	FALSE
limma_impute	lod_imputed	mVseto~9392	T_C_gv_WT	4.271	0.742	0.000	11.611	0.064	10.398	0.000	0.600	0.883	0.109	TRUE	TRUE
lm_impute	lod_imputed	mVseto~9392	T_C_gv_WT	4.271	0.742	0.000	11.531	0.064	10.380	0.000	0.599	0.884	0.103	TRUE	TRUE
limma_impute	lod_imputed	mVseto~9392	WT_KO_comp	0.511	0.462	0.002	5.145	0.090	10.398	0.000	0.263	0.661	0.109	TRUE	FALSE
lm_impute	lod_imputed	mVseto~9392	WT_KO_comp	0.511	0.462	0.001	5.463	0.085	10.380	0.000	0.274	0.649	0.103	TRUE	FALSE
limma_impute	lod_imputed	zvzYsk~2881	T_C_gv_KO	3.902	-0.004	0.982	-0.054	0.072	5.398	0.959	-0.185	0.177	0.107	TRUE	FALSE
lm_impute	lod_imputed	zvzYsk~2881	T_C_gv_KO	3.902	-0.004	0.982	-0.056	0.070	5.380	0.957	-0.179	0.171	0.104	TRUE	FALSE
limma_impute	lod_imputed	zvzYsk~2881	T_C_gv_WT	3.971	-0.133	0.178	-1.902	0.070	5.398	0.111	-0.310	0.043	0.107	FALSE	FALSE
lm_impute	lod_imputed	zvzYsk~2881	T_C_gv_WT	3.971	-0.133	0.172	-1.915	0.070	5.380	0.110	-0.309	0.042	0.104	FALSE	FALSE
limma_impute	lod_imputed	zvzYsk~2881	WT_KO_comp	-0.069	-0.129	0.360	-1.269	0.102	5.398	0.256	-0.386	0.127	0.107	TRUE	FALSE
lm_impute	lod_imputed	zvzYsk~2881	WT_KO_comp	-0.069	-0.129	0.339	-1.315	0.098	5.380	0.242	-0.377	0.118	0.104	TRUE	FALSE

8 Peptide-input facades

The mixed-effects lmer_nested facade and ropeca_nested require lower-level measurements below the analysis subject. The firth_nested facade is the peptide-input variant of firth shown above. All three still return protein-level contrasts.

fa_lmer_2f <- build_contrast_analysis(
  lfq_peptide_2f,
  "~ Subgroup + (1 | peptide_Id) + (1 | sampleName)",
  contrasts_2f,
  method = "lmer_nested"
)

fa_ropeca_2f <- build_contrast_analysis(
  lfq_peptide_2f,
  "~ Subgroup",
  contrasts_2f,
  method = "ropeca_nested"
)

fa_firth_peptide_2f <- build_contrast_analysis(
  lfq_peptide_2f,
  "~ Subgroup",
  contrasts_2f,
  method = "firth_nested"
)

results_peptide_2f <- bind_rows(
  fa_lmer_2f$get_contrasts(),
  fa_ropeca_2f$get_contrasts(),
  fa_firth_peptide_2f$get_contrasts()
) |>
  dplyr::select(dplyr::any_of(c(
    "modelName", "estimate_type", "protein_Id", "contrast", "avgAbd", "diff", "FDR",
    "statistic", "std.error", "df", "p.value", "conf.low", "conf.high",
    "sigma"
  ))) |>
  dplyr::mutate(
    significant = FDR < 0.1 & abs(diff) > 0.5
  )

results_peptide_2f |>
  dplyr::count(modelName, name = "n_results")

## # A tibble: 3 × 2
##   modelName     n_results
##   <chr>             <int>
## 1 firth_nested        240
## 2 lmer_nested         179
## 3 ropeca_nested       231

ggplot(results_peptide_2f, aes(x = diff, y = -log10(p.value), color = significant)) +
  geom_point(alpha = 0.6, size = 1.2) +
  facet_grid(contrast ~ modelName, scales = "free_y") +
  geom_vline(xintercept = c(-0.5, 0.5), linetype = "dashed", color = "grey60") +
  geom_hline(yintercept = -log10(0.1), linetype = "dashed", color = "grey60") +
  scale_color_manual(values = c(`TRUE` = "firebrick", `FALSE` = "grey70")) +
  labs(
    x = "log2 fold change",
    y = "-log10(p.value)",
    color = "FDR < 0.1\nand |diff| > 0.5"
  ) +
  theme_minimal(base_size = 12)

(#fig:volcano_peptide_2f)Volcano plots for the peptide-level facades. Rows are contrasts, columns are backends.

9 Remarks

The facades make it easy to benchmark alternative contrast backends without rewriting the analysis pipeline:

the protein-level facades now enforce aggregation before modelling
the peptide-level facades now explicitly require lower-level hierarchy below the analysis subject
the shared facade API still makes it straightforward to compare methods once the data level is chosen consistently

The results are comparable at the API level, but the comparison is only meaningful when methods using the same biological unit are plotted together.

10 Session Info

sessionInfo()

## R version 4.6.1 (2026-06-24)
## Platform: x86_64-pc-linux-gnu
## Running under: Ubuntu 24.04.5 LTS
## 
## Matrix products: default
## BLAS:   /home/biocbuild/bbs-3.24-bioc/R/lib/libRblas.so 
## LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.12.0  LAPACK version 3.12.0
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_GB              LC_COLLATE=C              
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
## 
## time zone: America/New_York
## tzcode source: system (glibc)
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
## [1] ggplot2_4.0.3    prolfqua_0.99.8  dplyr_1.2.1      BiocStyle_2.41.0
## 
## loaded via a namespace (and not attached):
##   [1] RColorBrewer_1.1-3     jsonlite_2.0.0         shape_1.4.6.1         
##   [4] magrittr_2.0.5         TH.data_1.1-5          estimability_2.0.0    
##   [7] jomo_2.7-6             magick_2.9.1           farver_2.1.2          
##  [10] logistf_1.26.1         nloptr_2.2.1           rmarkdown_2.32        
##  [13] GlobalOptions_0.1.4    vctrs_0.7.3            Cairo_1.7-0           
##  [16] minqa_1.2.8            tinytex_0.61           htmltools_0.5.9       
##  [19] forcats_1.0.1          progress_1.2.3         broom_1.0.13          
##  [22] mitml_0.4-5            sass_0.4.10            KernSmooth_2.23-27    
##  [25] bslib_0.12.0           htmlwidgets_1.6.4      plyr_1.8.9            
##  [28] sandwich_3.1-3         emmeans_2.0.4          plotly_4.12.1         
##  [31] zoo_1.9-1              cachem_1.1.0           mime_0.13             
##  [34] lifecycle_1.0.5        iterators_1.0.14       pkgconfig_2.0.3       
##  [37] Matrix_1.7-6           R6_2.6.1               fastmap_1.2.0         
##  [40] rbibutils_2.4.1        shiny_1.14.0           clue_0.3-68           
##  [43] digest_0.6.39          numDeriv_2016.8-1.1    colorspace_2.1-4      
##  [46] S4Vectors_0.51.10      crosstalk_1.2.2        labeling_0.4.3        
##  [49] httr_1.4.9             mgcv_1.9-5             compiler_4.6.1        
##  [52] withr_3.0.3            doParallel_1.0.17      S7_0.2.2              
##  [55] backports_1.5.1        UpSetR_1.4.1           pan_2.0               
##  [58] MASS_7.3-66            rjson_0.2.23           tools_4.6.1           
##  [61] otel_0.2.0             httpuv_1.6.17          nnet_7.3-21           
##  [64] glue_1.8.1             nlme_3.1-172           promises_1.5.0        
##  [67] grid_4.6.1             cluster_2.1.8.3        generics_0.1.4        
##  [70] operator.tools_1.6.3.1 gtable_0.3.6           formula.tools_1.7.1   
##  [73] tidyr_1.3.2            data.table_1.18.6.1    hms_1.1.4             
##  [76] utf8_1.2.6             BiocGenerics_0.59.12   ggrepel_0.9.8         
##  [79] foreach_1.5.2          pillar_1.11.1          limma_3.99.4          
##  [82] later_1.4.8            circlize_0.4.18        splines_4.6.1         
##  [85] lattice_0.23-1         survival_3.8-12        tidyselect_1.2.1      
##  [88] ComplexHeatmap_2.29.0  knitr_1.52             reformulas_0.4.4      
##  [91] gridExtra_2.3.1        bookdown_0.48          IRanges_2.47.5        
##  [94] stats4_4.6.1           xfun_0.62              statmod_1.5.2         
##  [97] matrixStats_1.5.0      stringi_1.8.9          yaml_2.3.12           
## [100] boot_1.3-32            evaluate_1.0.5         codetools_0.2-20      
## [103] Rfit_0.27.0            tibble_3.3.1           BiocManager_1.30.27   
## [106] cli_3.6.6              rpart_4.1.27           xtable_1.8-8          
## [109] Rdpack_2.6.6           jquerylib_0.1.4        dichromat_2.0-1       
## [112] Rcpp_1.1.2             coda_0.19-4.1          png_0.1-9             
## [115] parallel_4.6.1         prettyunits_1.2.0      lme4_2.1-0            
## [118] glmnet_5.2             viridisLite_0.4.3      limpa_1.5.2           
## [121] mvtnorm_1.4-2          lmerTest_3.2-1         scales_1.4.0          
## [124] purrr_1.2.2            crayon_1.5.3           GetoptLong_1.1.1      
## [127] rlang_1.3.0            multcomp_1.4-32        mice_3.19.0

Two-factor designs

2026-11-04

Contents

1 Purpose

2 Simulate two-factor data and encode subgroups

3 Define contrasts

4 Protein-input facades

5 Protein-level volcano comparison

5.1 Standard facades

5.2 Imputation and missingness facades

6 Looking at the strongest protein-level hits

7 Proteins that could not be estimated

7.1 Per-sample intensities of the missing proteins

7.2 Estimates for the missing proteins from `lm_impute` and `limma_impute`

8 Peptide-input facades

9 Remarks

10 Session Info

Two-factor designs

2026-11-04

Contents

1 Purpose

2 Simulate two-factor data and encode subgroups

3 Define contrasts

4 Protein-input facades

5 Protein-level volcano comparison

5.1 Standard facades

5.2 Imputation and missingness facades

6 Looking at the strongest protein-level hits

7 Proteins that could not be estimated

7.1 Per-sample intensities of the missing proteins

7.2 Estimates for the missing proteins from lm_impute and limma_impute

8 Peptide-input facades

9 Remarks

10 Session Info

7.2 Estimates for the missing proteins from `lm_impute` and `limma_impute`