Benchmark: Benchmark R6 class

In wolski/prolfqua: Proteomics Label Free Quantification

Benchmark R6 class

Description

Benchmark R6 class

Benchmark R6 class

Public fields

.data

data.frame

is_complete

todo

contrast

column name

toscale

which columns to scale

avgInt

average Intensity

fcestimate

estimate column

benchmark

todo

model_description

describe model

model_name

model description

hierarchy

todo

smc

summarize missing contrasts

summarizeNA

statistic to use for missigness summarization (e.g. statistic, or p-value)

confusion

todo

species

todo

FDRvsFDP

todo

Methods

Public methods

• Benchmark$new()

• Benchmark$data()

• Benchmark$missing_contrasts()

• Benchmark$complete()

• Benchmark$.get_confusion()

• Benchmark$get_confusion_benchmark()

• Benchmark$n_confusion_benchmark()

• Benchmark$plot_ROC()

• Benchmark$pAUC_summaries()

• Benchmark$pAUC()

• Benchmark$get_confusion_FDRvsFDP()

• Benchmark$n_confusion_FDRvsFDP()

• Benchmark$plot_FDRvsFDP()

• Benchmark$plot_score_distribution()

• Benchmark$plot_scatter()

• Benchmark$plot_precision_recall()

• Benchmark$clone()

---

Method new()

create Benchmark

Usage

Benchmark$new(
  data,
  toscale = c("p.value"),
  fcestimate = "diff",
  avgInt = "avgInt",
  benchmark = list(list(score = "diff", desc = TRUE), list(score = "statistic", desc =
    TRUE), list(score = "scaled.p.value", desc = TRUE)),
  FDRvsFDP = list(list(score = "FDR", desc = FALSE)),
  model_description = "protein level measurments, linear model",
  model_name = "medpolish_lm",
  contrast = "contrast",
  species = "species",
  hierarchy = c("protein_Id"),
  summarizeNA = "statistic"
)

Arguments

data

data.frame

toscale

columns ot scale

fcestimate

column with fold change estimates

avgInt

average protein/peptide/metabolite intensity

benchmark

columns to benchmark

FDRvsFDP

score for which to generate FDR vs FDP

model_description

describe model

model_name

model name

contrast

contrast

species

species (todo rename)

hierarchy

e.g. protein_Id

summarizeNA

examine this column to determine the proportion of missing values default statistic

columns

to create FPR vs FDP analysis for

---

Method data()

get data

Usage

Benchmark$data()

Returns

data.frame

---

Method missing_contrasts()

summarize missing contrasts

Usage

Benchmark$missing_contrasts()

Returns

data.frame

---

Method complete()

set or get complete. If true only proteins for which all contrasts are determinable are examined.

Usage

Benchmark$complete(value)

Arguments

value

TRUE if data should be complete (no missing contrasts)

---

Method .get_confusion()

get confusion data

Usage

Benchmark$.get_confusion(arrange)

Arguments

arrange

todo

---

Method get_confusion_benchmark()

get FDR summaries

Usage

Benchmark$get_confusion_benchmark()

---

Method n_confusion_benchmark()

nr of elements used to determine ROC curve

Usage

Benchmark$n_confusion_benchmark()

---

Method plot_ROC()

plot FDR summaries

Usage

Benchmark$plot_ROC(xlim = 0.5)

Arguments

xlim

limit x axis

Returns

ggplot

---

Method pAUC_summaries()

AUC summaries

Usage

Benchmark$pAUC_summaries()

---

Method pAUC()

AUC summaries as table

Usage

Benchmark$pAUC()

---

Method get_confusion_FDRvsFDP()

FDR vs FDP data

Usage

Benchmark$get_confusion_FDRvsFDP()

---

Method n_confusion_FDRvsFDP()

nr of elements used to determine ROC curve

Usage

Benchmark$n_confusion_FDRvsFDP()

---

Method plot_FDRvsFDP()

plot FDR vs FDP data

Usage

Benchmark$plot_FDRvsFDP()

Returns

ggplot

---

Method plot_score_distribution()

plot distributions of scores

Usage

Benchmark$plot_score_distribution(score)

Arguments

score

the distribution of which scores to plot (list)

Returns

ggplot

---

Method plot_scatter()

plot intensity vs scores

Usage

Benchmark$plot_scatter(score)

Arguments

score

the distribution of which scores to plot (list)

Returns

ggplot

---

Method plot_precision_recall()

plot precision vs recall

Usage

Benchmark$plot_precision_recall(precision_lim = 0.7, recall_lim = 1)

Arguments

precision_lim

limit shown precision from

recall_lim

limit shown recall to

Returns

ggplot

---

Method clone()

The objects of this class are cloneable with this method.

Usage

Benchmark$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

See Also

Other benchmarking: INTERNAL_FUNCTIONS_BY_FAMILY, ionstar_bench_preprocess(), make_benchmark(), ms_bench_add_scores(), ms_bench_auc()

Examples

dd <- dplyr::filter(prolfqua_data('data_benchmarkExample'), !is.na(statistic))
dd <- dd |> dplyr::mutate(avgInt = (c1 + c2)/2)
ttd <- ionstar_bench_preprocess(dd)
medpol_benchmark <- make_benchmark(ttd$data,
benchmark = list(
list(score = "estimate", desc = TRUE),
list(score = "statistic", desc = TRUE),
list(score = "scaled.p.value", desc = TRUE)
),
    fcestimate = "estimate",
    model_description = "med. polish and lm. density",
    model_name = "prot_med_lm"
)
medpol_benchmark$plot_score_distribution(list(list(score = "estimate", xlim = c(-1,2) ),
 list(score = "statistic", xlim = c(-3,10) )))
medpol_benchmark$get_confusion_benchmark()

#Benchmark$debug("plot_score_distribution")
benchmark <- make_benchmark(
  ttd$data,
  toscale =  c("moderated.p.value", "moderated.p.value.adjusted"),
  fcestimate = "estimate",
  benchmark = list(list(score = "estimate", desc = TRUE),
                   list(score = "statistic", desc = TRUE),
                   list(score = "scaled.moderated.p.value", desc = TRUE),
                   list(score = "scaled.moderated.p.value.adjusted", desc = TRUE)
  ),
  FDRvsFDP =
    list(list(score = "moderated.p.value", desc = FALSE),
         list(score = "moderated.p.value.adjusted", desc = FALSE)),
  model_description = "protein level measurments, lm model",
  model_name = "prot_lm"
)

bb <- benchmark$pAUC_summaries()
benchmark$complete(FALSE)
benchmark$smc$summary
benchmark$plot_score_distribution(list(list(score = "estimate", xlim = c(-1,2) ),list(score = "statistic", xlim = c(-3,10) )))
benchmark$plot_score_distribution()


bb <- benchmark$get_confusion_FDRvsFDP()
xb <- dplyr::filter(bb, contrast ==  "dilution_(4.5/3)_1.5")
bb <- benchmark$get_confusion_benchmark()


benchmark$plot_ROC(xlim = 0.1)
benchmark$plot_precision_recall()

benchmark$plot_FDRvsFDP()
benchmark$plot_scatter(list(list(score = "estimate", ylim = c(-1,2) ),list(score = "statistic", ylim = c(-3,10) )))
benchmark$complete(FALSE)
benchmark$missing_contrasts()
stopifnot(nrow(benchmark$pAUC_summaries()$ftable$content) == 4 * (4 + 1))
benchmark$complete(TRUE)
stopifnot(nrow(benchmark$pAUC_summaries()$ftable$content) == 4 * (4+1))
missum <- benchmark$missing_contrasts()$summary
stopifnot(nrow(missum) == 4)
stopifnot(ncol(missum) == 2)
# returns number of statistics
stopifnot(nrow(benchmark$n_confusion_benchmark()) == 4 * (4 + 1))
stopifnot(nrow(benchmark$n_confusion_FDRvsFDP()) == 2 * (4 + 1))
benchmark$pAUC()

wolski/prolfqua documentation built on April 27, 2024, 4:09 p.m.