tidyTPP provides an analysis pipeline for Thermal Protein Profiling (TPP) proteomics data, with functions for reading in data from protein quantification software, normalising data, analysis, and hit-finding.
The goal of tidyTPP is to provide a range of functions, each of which requires data in a straightforward data frame or tibble and returns long-format data in a tibble that are readily understood and easily used for further analysis. The style of data is inspired by the tidyverse, and tidyTPP makes use of several tidyverse packages - in particular tibble and ggplot2.
Importing:
-
Importing from Thermo Proteome Discoverer
-
Importing from Spectronaut
-
Importing with custom settings
Normalisation and Analysis:
-
Single- and multi-threaded fitting of temperature-dependent protein melting curves (TPP-TR)
-
Significance scores by melting point difference (
$\Delta T_m$ )1 -
Significance scores by non-pararametric analysis of response curves (NPARC)2
Hit-finding
-
Identifying hits by
$\Delta T_m$ p-value -
Identifying hits by NPARC F-score and p-value
Plotting
- Plotting TPP-TR melting-curves with ggplot2
You can install the development version of tidyTPP from GitHub with:
# install.packages("pak")
pak::pak("jackrogan/tidyTPP")A version of the entire pipeline, with defaults to match common uses, is available with apply_TPP_pipeline. This executes the following sequence:
import_TPP() |>
normalise_TPP() |>
analyse_TPP() |>
export_TPP() |>
get_TPP_hits()The function uses the following defaults in addition to the default behaviour of each included function:
-
Import supported proteomics quantification data formats
-
Calculate NPARC and
$\Delta T_m$ significance scores -
Report all hits with:
-
Adjusted NPARC-derived p-value < 0.05
-
$\Delta T_m$ in the same direction -
$\Delta T_m$ vs. control >$\Delta T_m$ between controls -
Full results table and identified hits are exported as .xlsx files with input file names modified with “_Results.xlsx” and “_Hits.xlsx” respectively, in the same location as the initial file input.
-
Usage:
# Example pipeline call
apply_TPP_pipeline(
datafile = "path_to_data_input.csv",
config = "path_to_config_file.csv",
path = "[optional]_path_to_shared_dir",
import_format = "format_name",
to_plot = FALSE,
max_cores = 4)The character argument import_format defines which of the inbuilt import functions to use to read in data files. Currently:
-
“Spectronaut”, “SN” - imports from peptide tables exported from Spectronaut DIA quantification reports.
-
“ProteomeDiscoverer”, “PD” - imports from protein tables exported from Thermo Proteome Discoverer DDA quantification results.
-
Default “Spectronaut”.
The (boolean) argument to_plot defines whether to show automated plots across all methods; if TRUE, normalisation, NPARC score distribution, and hit melting-point curves will all be plotted.
- Default FALSE.
The (integer) argument max_cores is passed to curve-fitting methods and defines maximum parallel cores to use for these operations.
- Default 4.
# Pipeline example: 20-protein test data
library(tidyTPP)
twenty_prot_report <-
system.file("extdata", "20_protein_peptide_report.csv", package = "tidyTPP")
experiment_config <-
system.file("extdata", "20_protein_config.csv", package = "tidyTPP")
# Apply pipeline
TPP_hits <-
apply_TPP_pipeline(datafile = twenty_prot_report,
config = experiment_config,
import_format = "spectronaut",
to_plot = TRUE,
max_cores = 4)Result:
TPP_hits
#> # A tibble: 2 × 10
#> Protein_ID Condition Comparison F_scaled p_adj_NPARC max_adj_pvalue
#> <chr> <chr> <chr> <dbl> <dbl> <dbl>
#> 1 Protein_I Treated Treated_vs_Control 23.2 0.0000901 0.0000141
#> 2 Protein_C Treated Treated_vs_Control 14.1 0.00118 0.0981
#> # ℹ 4 more variables: mean_melt_point <dbl>, mean_control_melt_point <dbl>,
#> # mean_diff_melt_point <dbl>, mean_control_diff_melt_point <dbl>
This example will walk through the main functions using 20-protein example data
import_ functions read two files - quantification results output in a program-specific format, and a configuration file, defining conditions, replicates and temperatures, in the format shown.
# 20-protein example data
twenty_prot_report <-
system.file("extdata", "20_protein_peptide_report.csv", package = "tidyTPP")
experiment_config <-
system.file("extdata", "20_protein_config.csv", package = "tidyTPP")
experiment_dir <- system.file("extdata", package = "tidyTPP")
# Config file contents
read.csv(experiment_config)[1:10,]
#> Experiment Condition Replicate Temp
#> 1 1 Control 1 37
#> 2 2 Control 1 41
#> 3 3 Control 1 44
#> 4 4 Control 1 47
#> 5 5 Control 1 50
#> 6 6 Control 1 53
#> 7 7 Control 1 56
#> 8 8 Control 1 59
#> 9 9 Control 1 63
#> 10 10 Control 1 67
# Import using spectronaut import format
twenty_prot_quan_data <-
import_spectronaut(datafile = "20_protein_peptide_report.csv",
config = "20_protein_config.csv",
path = experiment_dir)
#>
#> --------------------
#> TPP Data Import
#> --------------------
#> Read in:
#> 20_protein_peptide_report.csv
#> 20_protein_config.csv
#> --------------------
#> Pivoting to long table...
#> Transforming experiment names...
#> Matching to experiment config data...
#> Finding relative quantity values...
#> Data imported.
#> Found 20 proteins.
#> --------------------
# Resulting tibble
twenty_prot_quan_data
#> # A tibble: 800 × 8
#> Protein_ID Pep_N Match_N Condition Replicate Temp rel_quantity raw_quantity
#> <chr> <dbl> <dbl> <chr> <chr> <int> <dbl> <dbl>
#> 1 Protein_A 29 42 Control 01 37 1 1401047.
#> 2 Protein_A 29 42 Control 01 41 1.05 1464157.
#> 3 Protein_A 29 42 Control 01 44 1.06 1485400.
#> 4 Protein_A 29 42 Control 01 47 1.12 1568061.
#> 5 Protein_A 29 42 Control 01 50 1.35 1889459
#> 6 Protein_A 29 42 Control 01 53 1.29 1801219.
#> 7 Protein_A 29 42 Control 01 56 0.450 630379.
#> 8 Protein_A 29 42 Control 01 59 0.116 162054.
#> 9 Protein_A 29 42 Control 01 63 0.0945 132437.
#> 10 Protein_A 29 42 Control 01 67 0.166 232101.
#> # ℹ 790 more rowsData at this stage can be plotted to directly observe relative but not yet normalised protein curves.
plot_melt(twenty_prot_quan_data)
normalise_TPP transforms relative TPP-TR relative intensity data, normalising against fitted median melting curves, as described by Savitsky et al. 2014.1
# Normalise twenty-protein data, with visualisation
twenty_prot_normalised <-
normalise_TPP(TPP_tbl = twenty_prot_quan_data,
to_plot = TRUE)
#> --------------------
#> TPP Normalisation
#> --------------------
#> Quality Criteria:
#> col lower upper
#> 1 Pep_N 2 Inf
#>
#> jointP contains 20 Proteins.
#>
#> normP criteria:
#> Temp lower upper
#> 1 56 0.4 0.6
#> 2 63 -Inf 0.3
#> 3 67 -Inf 0.2
#>
#> normP contains 6 Proteins.
#> --------------------
#> Fit melting curve to normP medians:
#>
#> Fitting 4 protein curves...
#> Estimated total process time: 2.24 s
#> |== | 1 of 4|===== | 2 of 4|======= | 3 of 4|==========| 4 of 4
#> 4 of 4 fitted successfully.
#>
#> Total elapsed time: 0.03 s
#>
#> Best fitted normP median curve:
#> Condition Replicate R_sq
#> 1 Control 01 0.995
#> --------------------
#> 20 proteins normalised
#> --------------------
Resulting in the normalised data:
# Plotted melting data points
plot_melt(twenty_prot_normalised)
#> # A tibble: 800 × 9
#> Condition Replicate Temp Protein_ID Pep_N Match_N rel_quantity raw_quantity
#> <chr> <chr> <int> <chr> <dbl> <dbl> <dbl> <dbl>
#> 1 Control 01 37 Protein_A 29 42 1 1401047.
#> 2 Control 01 37 Protein_B 23 39 1 18637136
#> 3 Control 01 37 Protein_C 36 62 1 7913466.
#> 4 Control 01 37 Protein_D 76 108 1 1695097.
#> 5 Control 01 37 Protein_T 19 29 1 5209475
#> 6 Control 01 37 Protein_Q 14 17 1 2789951.
#> 7 Control 01 37 Protein_N 19 35 1 997140.
#> 8 Control 01 37 Protein_K 13 15 1 6945545
#> 9 Control 01 37 Protein_H 14 26 1 2126099
#> 10 Control 01 37 Protein_E 10 12 1 351740.
#> # ℹ 790 more rows
#> # ℹ 1 more variable: norm_coefficient <dbl>
analyse_TPP fits sigmoidal melting curves are fitted to each unique
combination of protein, condition and replicate, and features of the
curve (including
-
FDR-adjusted p-values are calculated for melting point differences from control conditions (
$\Delta T_m$ ) for each replicate1 -
Scaled F-scores and FDR-adjusted p-values are calculated from NPARC analysis for each protein2
Default behaviour is to apply both analyses:
# Analyse twenty-protein data
twenty_prot_analysed <-
analyse_TPP(TPP_tbl = twenty_prot_normalised,
control_name = "Control",
p_value_methods = c("melting_point", "NPARC"))
# Analysed data and fitted curves
plot_melt(twenty_prot_analysed)
#> Warning: Removed 1000 rows containing missing values or values outside the scale range
#> (`geom_segment()`).
#> Warning: Removed 3000 rows containing missing values or values outside the scale range
#> (`geom_segment()`).
#> Warning: Removed 4000 rows containing missing values or values outside the scale range
#> (`geom_segment()`).
All statistics are appended as new measurements to the tibble:
# Full table
twenty_prot_analysed
#> # A tibble: 800 × 25
#> Protein_ID Condition Replicate Temp Pep_N Match_N rel_quantity raw_quantity
#> <chr> <chr> <chr> <int> <dbl> <dbl> <dbl> <dbl>
#> 1 Protein_A Control 01 37 29 42 1 1401047.
#> 2 Protein_A Control 01 47 29 42 1.15 1568061.
#> 3 Protein_A Control 01 50 29 42 1.30 1889459
#> 4 Protein_A Control 01 59 29 42 0.0904 162054.
#> 5 Protein_A Control 01 67 29 42 0.166 232101.
#> 6 Protein_A Control 01 53 29 42 1.23 1801219.
#> 7 Protein_A Control 01 56 29 42 0.450 630379.
#> 8 Protein_A Control 01 44 29 42 1.03 1485400.
#> 9 Protein_A Control 01 63 29 42 0.131 132437.
#> 10 Protein_A Control 01 41 29 42 0.999 1464157.
#> # ℹ 790 more rows
#> # ℹ 17 more variables: norm_coefficient <dbl>, F_scaled <dbl>,
#> # p_adj_NPARC <dbl>, a <dbl>, b <dbl>, plateau <dbl>, melt_point <dbl>,
#> # infl_point <dbl>, slope <dbl>, R_sq <dbl>, Comparison <chr>,
#> # diff_melt_point <dbl>, min_comparison_slope <dbl>, min_R_sq <dbl>,
#> # min_slope <dbl>, max_control_plateau <dbl>, adj_pvalue <dbl>
# Statistics only
one_prot_stats <-
twenty_prot_analysed |>
dplyr::filter(Protein_ID == "Protein_A", Condition != "Control") |>
dplyr::select(Protein_ID,
Condition,
Replicate,
F_scaled,
p_adj_NPARC,
melt_point,
diff_melt_point,
adj_pvalue
) |>
dplyr::distinct()
# 1. Melting point
one_prot_stats |>
dplyr::select(Protein_ID,
Condition,
Replicate,
melt_point,
diff_melt_point,
adj_pvalue
)
#> # A tibble: 2 × 6
#> Protein_ID Condition Replicate melt_point diff_melt_point adj_pvalue
#> <chr> <chr> <chr> <dbl> <dbl> <dbl>
#> 1 Protein_A Treated 01 56.0 0.0801 0.998
#> 2 Protein_A Treated 02 56.0 0.0851 0.998
# 2. NPARC
one_prot_stats |>
dplyr::select(Protein_ID,
Condition,
F_scaled,
p_adj_NPARC,
) |>
dplyr::distinct()
#> # A tibble: 1 × 4
#> Protein_ID Condition F_scaled p_adj_NPARC
#> <chr> <chr> <dbl> <dbl>
#> 1 Protein_A Treated 0.128 0.994get_TPP_hits filters and summarises analysed data. Hits, by default,
have an NPARC-derived, FDR-adjusted p-value below
0.05,2
# Get hits from analysed twenty-protein data
twenty_prot_hits <-
get_TPP_hits(TPP_data = twenty_prot_analysed,
hit_criteria = "default_hit_criteria",
to_plot = TRUE,
annotate = "melt_point")
#> --------------------
#> TPP Hit Identification
#> --------------------
#> Hit Criteria:
#> NPARC_pvalue_threshold DTm_same_sign DTm_gt_Dcontrol
#> 0.05 TRUE TRUE
#> --------------------
#> Exporting hit data...
#> --------------------
#> TPP Export
#> --------------------
#> Export format: xlsx
#> Saving TPP_hits.xlsx ...
#> Saved.
#> --------------------
#> Plotting hit melting curves...
#> 2 hits found.
#> --------------------
twenty_prot_hits
#> # A tibble: 2 × 10
#> Protein_ID Condition Comparison F_scaled p_adj_NPARC max_adj_pvalue
#> <chr> <chr> <chr> <dbl> <dbl> <dbl>
#> 1 Protein_I Treated Treated_vs_Control 23.2 0.0000901 0.0000141
#> 2 Protein_C Treated Treated_vs_Control 14.1 0.00118 0.0981
#> # ℹ 4 more variables: mean_melt_point <dbl>, mean_control_melt_point <dbl>,
#> # mean_diff_melt_point <dbl>, mean_control_diff_melt_point <dbl>
get_TPP_hits exported hit data by default, and full data can also be exported as an excel .xlsx spreadsheet with export_TPP. Alternatives are delimited text files (.csv, .tsv) or R data.
# Export as .xlsx
export_TPP(TPP_data = twenty_prot_analysed,
file_name = "TPP_results.xlsx",
format = "xlsx")1. Savitski, M. M. et al. Tracking cancer drugs in living cells by thermal profiling of the proteome. Science 346, (2014).
2. Childs, D. et al. Nonparametric analysis of thermal proteome profiles reveals novel drug-binding proteins*. Molecular & Cellular Proteomics 18, 2506–2515 (2019).