try_query() now correctly handles errors that don’t
return a response object. We also handle gzip decompression problems
better since some databases compressed responses were not handled
correctly.calculate_go_enrichment() got additional arguments.
replace_long_name: a logical argument that specifies if
GO term names above 50 characters should be replaced by the GO ID
instead for the plot. This ensures that the plotting area doesn’t become
too small due to the long name. The default is TRUE.label_move_frac: a numeric argument between 0 and 1
that specifies which labels should be moved outside of the bar. The
default is 0.2, which means that the labels of all bars that have a size
of 20% or less of the largest bar are moved to the right of the bar.
This prevents labels from overlapping with the bar boundaries.fetch_alphafold_aligned_error(),
fetch_alphafold_prediction(), fetch_mobidb(),
fetch_quickgo(), fetch_uniprot() and
fetch_uniprot_proteome() got additional arguments:timeout: a numeric value specifying the time in seconds
until the download times out.max_tries: a numeric value that specifies the number of
times the function tries to download the data in case an error
occurs.min_n_peptides parameter to the
calculate_protein_abundance() function. This allows users
to specify the minimum number of peptides per protein needed for
analysis. Default is set at three peptides.fetch_uniprot() previously had an issue where it
incorrectly identified certain IDs as UniProt IDs, such as ENSEMBL IDs.
For example, it would incorrectly interpret
"CON_ENSEMBL:ENSBTAP00000037665" as "P00000".
To address this, the function now requires that UniProt IDs are not
preceded or followed by letters or digits. This means that UniProt IDs
should be recognized only if they stand alone or are separated by
non-alphanumeric characters. For instance, in the string
"P02545;P20700", both "P02545" and
"P20700" are correctly identified as UniProt IDs because
they are separated by a semicolon and not attached to any other letters
or digits. Fixes issue #245.calculate_go_enrichment() now correctly uses the total
number of provided proteins for the contingency table. Previously it
falsely only considered proteins with a GO annotation for the enrichment
analysis.fetch_uniprot() and
fetch_uniprot_proteome() are more resistant to database
connection issues. They also give more informative messages as to why
the data could not be retrieved. Fixes issue #252.qc_csv() now properly works if the column supplied to
the condition argument is a factor. Fixes issue #254.analyse_functional_network() function now includes
enhanced error handling to ensure it fails gracefully in case of any
issues. Fixes issue #259.version parameter for
analyse_functional_network() has been updated to 12.0,
aligning with the latest STRINGdb version. Fixes issue #244.calculate_treatment_enrichment() received additional
arguments.
fill_colours: a character value that can be used to
provide custom colours to the plot.fill_by_group: a logical value that specifies if the
bars in the plot should be filled according to group.facet_n_col: specifies the number of columns in the
facet plot if a group column was provided.calculate_go_enrichment() got additional arguments.
facet_n_col: determines the number of columns the
faceted plot should have if a group column is provided.plot_title: specifies the title of the plot.min_n_detected_proteins_in_process: argument for
plotting that specifies the minimum number of proteins a GO term needs
to be detected for.enrichment_type: specifies what kind of enrichment
should be calculated. It can be “all”, “enrichment” or “deenrichment”.
This argument affects how the fisher.test() calculates the
enrichment. A two-sided test will be used for “all”, while a one-sided
test in the specific direction will be used for “enriched” or
“deenriched”.barplot_fill_colour: specifies the colours used to fill
the bars in the barplot. Needs always at least two values one for
deenriched the other for enriched.plot_style: We added a new plot type to the function.
The standard plot is still the default and is called “barplot”, while
the new plot type is “heatmap”. The heatmap plot is especially useful
for comparing GO enrichments of multiple groups.heatmap_fill_colour: specifies the colours used for the
colour gradient of heatmap plots.heatmap_fill_colour_rev: a logical value that specifies
if the colour gradient should be reversed.plot_cutoff: is now more flexible. You can provide any
number with the “top” cutoff. E.g. “top10”, “top5”.barcode_plot() received additional arguments.
facet_n_col: determines the number of columns the
faceted plot.fill_colour_gradient: specifies the colours used for
the colour gradient if the colouring column is
continous.fill_colour_discrete: specifies the colours used for
the fill colours if the colouring column is discrete.mako_colours to the package that contain 256
colours of the “mako” colour gradient.drc_4p_plot() received additional arguments.
facet_title_size: determines the size of the facet
titles.export_height: determines the output height of an
exported plot in inches.export_width: determines the output width of an
exported plot in inches.x_axis_limits: user supplied x-axis limits for each
plot.colours: determines colours used for the plot.fit_drc_4p() and parallel_fit_drc_4p()
have been updated in the latest version of protti,
leading to slight adjustments in their computational results compared to
previous versions.
anova_cutoff lets you define the ANOVA adjusted p-value
cutoff (default 0.05).n_replicate_completeness replaces
replicate_completeness. Now we encourage you to provide a
discrete number of minimal replicates instead of a fraction that is
multiplied with the total number of replicates. This is particularly
important to ensure that thresholds between different datasets and data
completeness levels are reproducible.n_condition_completeness replaces
condition_completeness. Same as above, we encourage you to
provide the minimal number of conditions that need to meet the replicate
completeness criteria as a number instead of a fraction.complete_doses is a new optional argument that should
be provided if the dataset is small and potentially incomplete. This
ensures that no matter if any doses are missing from the provided data
or not, the MNAR of the curve is calculated correctly. We would
recommend always providing it to ensure proper reproducibility.dose_MNAR
column are now part of the hits. To get back to the old output you can
just exclude them again from the ranked results.filter argument works. This argument controls if
significance statistics should be annotated in the data.
"pre": This previously filtered curves by the
completeness as well as the ANOVA adjusted p-value prior to fitting
curves. Now it only filters by completeness. This also allows it to be
an option for the parallel_fit_drc_4p() function."post": Is still the default value and still just
annotates the data without any filtering."pre" to remove
usually not trustworthy features with too few complete concentrations
from the data before p-value adjustment and curve fittings. This will
solidify your confidence that features without a dose-response behavior
are true negative. The point is that it is better to not include any
features with too few values because they are potentially false
negative.normalise() now correctly works with grouped data.
Previously it would only correctly work with ungrouped data frames. Now
you can group the data to calculate group specific normalisations. If
you want to compute a global normalisation for the dataset, you need to
ungroup the data before using the function as usual. This fixes issue
#209.qc_sequence_coverage() now correctly displays medians
in faceted plot. This fixes issue #202 and #213.fit_drc_4p() and parallel_fit_drc_4p() now
correctly calculates the ANOVA p-value. Previously the number of
observations for each concentration was not provided correctly.fetch_uniprot() now correctly retrieves information if
an input ID was also part of a non-conform input ID combination. When
e.g. c("P02545", "P02545;P20700") was provided, previously
the "P02545" accession was dropped from the
input_id column even though it is also present on its own
and not only in combination with "P20700". The new output
now contains 3 rows, one for each ID, with "P02545" having
one row with the input_id `"P02545" and one
with the input_id "P02545;P20700". This also
means that the input_id column now always contains the
provided input IDs and not only if they were non-conform input ID
combinations.fit_drc_4p() and parallel_fit_drc_4p()
the arguments replicate_completeness and
condition_completeness are now deprecated. Please use
n_replicate_completeness and
n_condition_completeness instead.qc_charge_states(),
qc_peptide_type() and qc_missed_cleavages().
Also made appearance more uniform between methods "count"
and "intensity".fetch_uniprot() now returns nothing instead of a
partial output if some of the requested batches could not be retrieved
due to database issues (e.g. timeout because of too many requests). This
addresses issue #203, which requests this change, because the warning
message regarding the partial output can be easily missed and users
might wrongfully assume that all information was retrieved successfully
from UniProt.find_peptide() now preserves the groups of the original
data. This does not affect any of the calculations.calculate_sequence_coverage() now works on grouped
data.correct_lip_for_abundance() was added. It corrects
LiP-peptides for changes in protein abundance and calculates their
significance using a t-test. The function is based on the MSstatsLiP
package developed by the Vitek Lab. Big thanks to @FehrAaron for implementing it!qc_cvs() received a new argument called
max_cv that specifies the maximum CV that should be
included in the plot.peptide_profile_plot() received a new argument called
complete_sample. If set to TRUE, each protein
gets assigned all sample names that are found in the input data. This
ensures that the plot always contains all samples on the x-axis even if
there are no measured intensities for a specific sample. The default is
FALSE, which is the original behaviour of the
function.volcano_plot() received the colour
argument that allows the user to provide custom colours for points.find_peptide() and
assign_peptide_type() by only computing on the smallest
possible subset of data before joining back to the original data
frame.calculate_treatment_enrichment() can now be applied on
data frames with multiple different groups. The enrichment will be
calculated for each group separately. If the data is plotted, each group
is displayed in a separate facet. The group is provided to the new
group argument.qc_pca(): If the condition argument is numeric a colour
gradient is used instead.volcano_plot() now also works interactively if there
are no significant hits.fetch_chebi(): fixed an issue caused by
na_if() that changed its behaviour after the recent
dplyr update.qc_proteome_coverage(): fixed the label order of
fractions of proteins detected and not detected in the proteome. Fixes
issue #194.calculate_protein_abundance() now correctly retains
columns if for_plot = TRUE. Previously the columns to
retain were not joined considering the precursor column, which lead to
duplications of information where it did not belong. Fixes issue
#197.fetch_kegg() now returns the pathway name correctly
again.qc_intensity_distribution(),
qc_median_intensities(), qc_charge_states(),
qc_contaminants(), qc_missed_cleavages(),
qc_peptide_type(), qc_ids(): If the provided
sample column is of type factor, the level order won’t be overwritten
anymore. *fit_drc_4p(): If there are no correlations an
empty data frame is returned to prevent errors in
parallel_fit_drc_4p().calculate_sequence_coverage() does not fail anymore if
a protein only contains NA peptide sequences.qc_sequence_coverage() does not return a plot anymore
if plot = FALSE. This fixes issue #207.qc_data_completeness() if sample was of type
factor the function did not properly facet the data when
the digestion argument was provided. Now we filter out all
0% completeness values that come from factor levels that are not present
in subsetted data.calculate_go_enrichment() can now be applied on data
frames with multiple different groups. The enrichment will be calculated
for each group separately. If the data is plotted, each group is
displayed in a separate facet. The group is provided to the new
group argument. The y_axis_free argument
determines if the y-axis of the faceted plot is “free” or “fixed”.version argument to
fetch_alphafold_prediction() that specifies which verison
of the database should be retrieved. The default is currently the newest
version "v4".qc_ranked_intensities() was added. It ranks protein,
peptide or precursor intensities from highest to lowest. Ranked
intensities can also be plotted using the plot
argument.fetch_chebi() recieved a timeout argument
that specifies after how many seconds the connection to the database
should timeout. The default is 60 seconds as previously used.pval_distribution_plot() facets now have the correct
style.calculate_protein_abundance() requires at least three
distinct peptides for quantification. The function now applies this rule
for each sample independently except for checking the whole dataset to
contain at least three distinct peptides.fetch_alphafold_aligned_error() was added. It fetches
the aligned error matrix for structure predictions from the AlphaFold
EBI database.predict_alphafold_domain() was added. It uses a
graph-based community clustering algorithm of AlphaFold predicted
aligned errors in order to infer protein domains in AlphaFold
predictions. The code is based on python code by
Tristan Croll.assign_missingness() now correctly deals with unequal
replicate numbers of comparisons. In addition there is a message
returned if an unequal number of replicates is detected for a
comparison.fetch_chebi() fixed a bug that prevented the function
from failing gracefully if there is a connection problem to the
server.extract_metal_binders() now checks if the provided data
frames are NULL. If yes, a message and NULL is
returned.fetch_mobidb() was updated after the API changed.fetch_alphafold_aligned_error() and
predict_alphafold_domain() functions.iq
package to protti.
calculate_protein_abundance() now has the method
"iq" again as an option.fetch_pdb() now also retrieves information on
engineered mutations, non-standard monomers, secondary structure and
binding interfaces of ligands.extract_metal_binders() was completely redone. This was
in response to the UniProt update and rework of the binding column
provided by UniProt. This function extracts and concatenates all metal
binding information available for a protein based on the UniProt and
QuickGO databases. Therefore, this function now also takes gene ontology
(GO) information from QuickGO as input. Instead of being able to provide
column names to specific argument of the function you now only provide
the data frames. This makes the function less flexible but reduces the
amount of arguments required to achieve the same result. You just need
to make sure that the input data frames contain columns with the correct
names as stated by the function documentation.fetch_quickgo() was added. It fetches gene ontology
(GO) information from the QuickGO EBI database. The retrieved
information can either be GO annotations for provided UniProt IDs or
Taxon identifiers, a list of all GO terms or a “slims” subset of GO IDs
that can be generated based on provided GO IDs.fetch_chebi() now has the stars argument
with which one can select the evidence levels for which entries should
be retrieved.auth_seq_id column that is part of the output
of the fetch_pdb() function. Previously, the column could
contain duplicated or missing positions. This was formerly identified by
comparing the number of positions within the auth_seq_id
column and the number of residues in the deposited
pdb_sequence. Positions are now correct. The original
output can be found in the auth_seq_id_original
column.calculate_diff_abundance() function the
intensity column can now be retained with the
retain_columns argument. This was previously not possible
until now since this column was used to reduce the annotation dataset.
However, after reassessing the benefit of this filter step, it seemed
not necessary.calculate_diff_abundance() that would not duplicate the
data. However, this seems not to be the case, which can lead to wrong
p-value adjustment. p-value adjustment was originally performed after
the columns indicated in retain_columns are joined back to
the data. Now p-value adjustment is performed prior to retaining columns
as well as only on the subset of data that actually contains p-values.
Previously we (by default filter_NA_missingness = TRUE)
only filtered out NAs in the missingness
column prior to p-value adjustment. However, it is possible that
missingness is not NA but the p-value is
NA. Now for all methods except for "proDA" we
remove NA p-values before p-value adjustment. For
"proDA" data is handled as previously since p-values are
never NA.fetch_pdb() was changed to 100
(from 200). This was done since more information is retrieved now, which
slows to function down and is slightly improved when batch sizes are
smaller.try_query() now only retries to retrieve information
once if the returned message was “Timeout was reached”. In addition, a
timeout and accept argument have been
added.fetch_uniprot() and fetch_uniprot_proteome()
function accordingly. Please be aware that some columns names might have
changed and your code might throw error messages if you did not adjust
it accordingly.map_peptides_on_structure() was still using
“residue” as its input. We now use
find_peptide_in_structure() to generate the correct input
column.fetch_uniprot() and
fetch_uniprot_proteome(). As UniProt has updated their
website and their programmatic access, we now download the information
from the legacy version temporarily. A real fix will follow.fetch_kegg(). The function did not
retrieve any data after the API URL had changed.fetch_eco() was added. It fetches evidence &
conclusion ontology information from the EBI database.qc_proteome_coverage() now has the
reviewed argument that specifies if only reviewed entries
in UniProt should be considered as the proteome. The default is
TRUE and stays the same as previously.volcano_plot() now has the facet_scales
argument that specifies if the scales should be “free” or “fixed” when a
faceted plot is created. The arguments that can be provided are the same
that can be provided to the scales argument of
ggplot2::facet_wrap(). The new default is now
"fixed".pval_distribution_plot() now has the optional
facet_by variable that allows faceting of the plot.map_peptides_on_structure() that caused
an error if the column provided to the auth_seq_id argument
was called “residue”.volcano_plot() that did not calculate
the horizontal cutoff line correctly if there were multiple significance
values that have the same adjusted significance value. Now it correctly
uses the two p-values closest to the cutoff for the line position
calculation. In addition, points were not correctly displayed if no
horizontal cutoff line was created due to no significant values. Now all
values are displayed correctly.try_query() function now
returns errors as a character string if it encounters one. Functions
using try_query() however, still expected NULL
if there was an error. Also adjusted additional fetch functions that do
not use try_query() to fail gracefully and to return
informative messages upon encountering errors.calculate_go_enrichment() now has the argument
label. If TRUE labels are added to the plot
that specify how many proteins from the specific GO term are among the
significant hits. This new argument is by default
TRUE.version argument in the
analyse_functional_network() function.protti depends on and not also
packages it suggests.iq package
from protti for now since iq is currently not
available on CRAN. Once it is available again we will add the
functionalities back.fetch_metal_pdb() now gives more informative feedback
regarding the reasons resources were not fetched correctly.qc_proteome_coverage that flipped the
“Detected” and “Not detected” labels.highlight argument of the
woods_plot() function was used huge plots were generated
when more than 20 proteins were provided to the function. This is fixed
and was due to a wrong variable used in a loop.create_structure_contact_map() now takes an optional
data2 argument in which a second data frame of positions
and structures can be provided. The positions in this data frame are
used for distance calculations relative to the positions provided in the
data argument. This helps to reduce the size of the map
since only comparisons of interest are made. If a selection provided
through the data argument should be compared to the whole
structure the data2 argument should not be provided, which
is the previous default setting.parallel_create_structure_contact_map() can create
structure contact maps using parallel processing. It is also recommended
for sequential processing if a large number of contact maps should be
created. The non-parallel function should not be used if more than 50
maps should be created at the same time. In order to reduce contact maps
to domains or even only peptides, a search pattern is created that
selects only relevant regions. This pattern is shared over all maps and
would become too large if too maps are created at once.qc_cvs()
function now has a grey colour. This ensures that the other colours for
each condition match the colours of other quality control plots.run_order argument of
qc_sample_correlation() now has a gradient colour scheme
that is easier to interpret than distinct colours. The colours are
inspired by the “plasma” colour scheme of the viridis package.qc_intensity_distribution() is of type factor instead of
character the provided factor levels are used for sample ordering in the
plot. This allows for custom sample ordering. If you want this
functionality in other functions, please let us know in an issue on
GitHub.qc_ids(), which caused an error when the
optional condition argument was not provided. Also fixed a
bug that did not take the state of the
remove_na_intensities argument into account.auth_seq_id variable in structure files was handled
as a numeric variable even though it sometimes can be character. This
was fixed in all functions using it. These include:
fetch_pdb_structure(),
fetch_alphafold_prediction() (output is now numeric),
fetch_pdb() (provides the auth_seq_id column
not as a list vector but as character vector with semicolons as
separators), find_peptide_in_structure() (now uses the new
auth_seq_id format and returns the character vector of all
positions of each peptide as auth_seq_id in additon to
auth_seq_id_start and auth_seq_id_end),
create_structure_contact_map(),
map_peptides_on_structure() (now take the new
auth_seq_id column instead of start and end positions as
input).map_peptides_on_structure(), which
caused an error when file names larger than 256 characters were
generated. This was the case if the number of proteins that are part of
one structure is very large (e.g. ribosome). If the number of proteins
is too large to fit into a normal length file name, they are abbreviated
as for example “51_proteins”.fetch_alphafold_prediction() and
fetch_pdb_structure() now return more informative messages
if individual IDs have not been retrieved correctly (e.g. internet
connection problem or outdated ID).calculate_diff_abundance(), comparison names starting with
numbers were not supported. This has been fixed. There are no more
restrictions for condition names.fetch_pdb() was added. It fetches PDB structure
metadata from RCSB.fetch_pdb_structure() was added. It fetches atom level
data for a PDB structure from RCSB.fetch_metal_pdb() was added. It fetches information
about protein-metal binding sites from the MetalPDB database.find_peptide_in_structure() was added. It returns the
positions of a protein region, peptide or amino acid within a protein
structure using UniProt positions as input. This is necessary because
often amino acid positions in a protein structure vary from their
positions in the UniProt protein sequence.map_peptides_on_structure() was added. It can map
peptides onto PDB structures based on their positions. This is
accomplished by replacing the B-factor information in the structure file
with values that allow highlighting peptides when the structure is
coloured by B-factor.create_structure_contact_map() was added. It creates a
contact map of a subset or of all atom or residue distances in a
structure file.calculate_aa_scores() was added. It calculates a score
for each amino acid of a protein based on the product of the
-log10(adjusted p-value) and the absolute log2 fold change per
peptide.woods_plot() recieved the highlight
argument, which allows to highlight peptides in the plot with an
asterisk based on a logical column. It also got export and
export_name arguments which now makes it possible to
directly export plots from the function. The new target
argument allows the user to specify one or multiple proteins from their
data frame for which a plot should be returned, however the default
option is to return plots for all proteins in the provided data frame.
Now it is also possible to provide more than 20 proteins at a time while
the facet argument is TRUE. For every 20
proteins a new plot is created and all of them are returned together in
a list.assign_missingness() and
calculate_diff_abundance() now also take "all"
as input to their ref_condition argument. This will create
all pairwise condition pairs. Previously only one reference condition
could be provided.volcano_plot() can now plot unadjusted p-values while
the horizontal cutoff line is based on the adjusted p-value. the
significance_cutoff argument was modified to also take a
second value which specifies the adjusted p-value column name. In that
case the y-axis of the plot could display p-values that are provided to
the significance argument, while the horizontal cutoff line
is on the scale of adjusted p-values transformed to the scale of
p-values. The provided vector can be
e.g. c(0.05, "adj_pval"). In that case the function looks
for the closest adjusted p-value above and below 0.05 and takes the mean
of the corresponding p-values as the cutoff line. If there is no
adjusted p-value in the data that is below 0.05 no line is displayed.
This allows the user to display volcano plots using p-values while using
adjusted p-values for the cutoff criteria. This is often preferred
because adjusted p-values are related to unadjusted p-values often in a
complex way that makes them hard to be interpret when plotted.Multiple functions have been renamed. More function follow the convention that they should be verbs. Other functions are more similar to each other in their naming. The old functions still work but they are deprecated and will be removed in the future. Please use the new versions instead.
diff_abundance() has been renamed to
calculate_diff_abundance().peptide_type() has been renamed to
assign_peptide_type().median_normalisation() has been renamed to
normalise(). The normalisation method is now defined
through an argument called method.sequence_coverage() has been renamed to
calculate_sequence_coverage().network_analysis() has been renamed to
analyse_functional_network().treatment_enrichment() has been renamed to
calculate_treatment_enrichment().go_enrichment() has been renamed to
calculate_go_enrichment().kegg_enrichment() has been renamed to
calculate_kegg_enrichment().volcano_protti() has been renamed to
volcano_plot().plot_drc_4p() has been renamed to
drc_4p_plot().plot_peptide_profiles() has been renamed to
peptide_profile_plot().plot_pval_distribution() has been renamed to
pval_distribution_plot().calculate_diff_abundance() had a bug that could not
correctly deal with condition names containing spaces when a moderated
t-test or the proDA algorithm was used. This has been fixed.scale_protti() can now deal with the case that a vector
of only equal numbers (e.g. c(1, 1, 1)) is provided. In
this case it always scales all values to 1 for
method = "01" and to 0 for
method = "center".pval_distribution_plot() had a bug centered each bin
around values, which meant that the first and last bin were smaller.
This has been fixed.header argument was removed from
try_query(). It can now be directly supplied to the elipsis
(…) of this function depending on which read method is used. For
type = "text/tab-separated-values" the argument
col_names of the corresponding readr function
read_tsv() can be supplied.split_metal_name() removes more non-metal specific
words that lead to false identifications. Also many metal names
containing a “-” are now considered. A bug was resolved that did not
deal with iron-sulfur cluster information correctly due to a wrong
variable name.extract_metal_binders() now also correctly identifies
monovalent inorganic cations as metals.passed_filter
example column that is used in functions afterwards.diff_abundance() function. Previously, the p-values in the
whole result table were adjusted together. Now adjustments are made by
comparisons. This change only affects analyses that have more than one
comparison (two conditions).fetch_uniprot(),
fetch_uniprot_proteome() and fetch_kegg().
They now fail gracefully with informative messages if there is no
internet connection, the database times out or the data resource has
changed. Specifically the underlying try_query() function
was changed.