- Fixed
NAMESPACE
exports for data objects - Added more readable default labels
- Fixed bug where the vignette could not be built
- Fixed bug where
max.genes
would not work for ggplot/long table format
- Added the
max.genes
option to plot genemaps -- by default it keeps all genes (so nothing changes from prior versions), but allows limiting their number
- Minor fixes in
show
methods - Now pathway metadata can be specified (optional), and the automatic
columns
mod
(oc or bin) andn.genes
(after mapping to human data) are added - Pathway metadata can be displayed in pathwaymaps by using
vars.meta.path
- Updates in vignette, including graphical scheme
- Updated syntax
- Added section on pathway/sig prioritisation
- Added prioritisers for pathways and signatures
- Fixed inexistent default argument in
get.tab.pathwaymap
- Legacy batch functions are skipped in unit testing
- Important change:
group
now belongs toist.results
objects, not toist.pathways
. This avoids having to fit pathways again to change the groups. - Error raises if
id.ref
orid.ist
are allFALSE
- The orthology mapping and the ensembl-symbol mapping are now accessible
with
::
(lazydata was not set in DESCRIPTION) ist.path
can be defined with empty pathway tables foroc
orbin
- Added getters for data, reponse, gene and pathway ids, sample groups and their levels
- Fixed bug in pheatmap with NULL title
- Updated NAMESPACE
- Allowed cartesian in data.table joins, as this would raise exceptions in when going to 100k+ rows in delta tables
- When
sig.ids
has no overlap with a pathway,get.tab.genemap()
would raise an error - now returns NULL with a message - Added option of plot titles, and genemaps have the pathway name by default
- Added helpers
save.pheatmap()
,save.genemaps()
andsave.boxplots()
to provide a default way of exporting the basic plots
- Minor modifications so that CI goes through
- Fixed issue of missing gene labels in
ist.results
- Added two label mapping vectors ready to use in
ist.results
instead of data.frame - Expanded options for
pheatmap
through theargs.pheatmap
argument and its defauls - Added
getGroups()
,getMetaSig()
and refactored some slot accessions - By default,
group
is ordered now
- Fix order issue in boxplots
- Fixed default boxplot: now shows original data groups
- Fixed colour scale in heatmaps
- Updated vignette
- Exported plotting functions
- Small fixes to
do.isx()
(legacy) and toplot.ist.boxplot()
(bug)
- Fixed error in default gene labels
- Now signatures can be subsetted in every plot type with
sig.ids
- Removed assertions of strict subsets in signature/pathway ids
- Added default orthology mappings
data.list.orth
and gene labels for plottingdata.dt.genelabels
- Added the slot
@vec.gene2label
toist.results
; if provided, gene labels are switched in the genemaps
- Fixed original signature metadata, that had extra rows
- Moved
id.ist
andid.ref
to theist.results
object - Updated vignette with current signature and pathway classes
- Updated CI dependencies
- Added more getters, esp
getSignatures()
andgetPathways()
- Added plotting functions:
plot.ist.boxplots()
,plot.ist.genemaps()
,plot.ist.pathwaymaps()
- The
group
field for reference data is optional. If omitted, grouping uses the differenty
values.
- Essential tables are now ready in
ist.results
objects - Getter functions (e.g.
getTabSignatures()
) ease slot accession inist.results
objects - Added helpers to generate the tables
- Added first getters to retrieve results tables (TODO: unit testing)
- Fixed minor bug in imports
- Added
ist.results
class for holding pathway results - Added
fit()
andshow()
methods
- Prepared parallel implementations of fitting and predicting
ist.pathways
- However, now the SerialParam backend is used due to MulticoreParam hanging for no apparent reason (probably missing exports)
ist.pathways
should accept aBiocParallel
backend now forfit()
andpredict()
- Added helper function with repeated logic for fitting
ist.pathways
- Added a helper with the logic of applying fold changes to data matrices, as it is shared between some classes
- Added
predict()
toist.signatures
- Added
fit()
andshow()
method forist.signatures
- Now
orthIDcon()
is a unique interface for mapping within and between species - Added
BiocParallel
dependency
- Small change in
sample.env.ist
- Added new class
ist.signatures
to hold signatures from varios organisms
- Added new sample data
sample.env.ist
, more similar to what we are actually using IST on. Contains signatures from several organisms with metadata and pathways with metadata.
- Added defaults to OC models (as they were present for BIN)
- Now OC should accept other kernels as well (untested)
- Added a slot for models, storing the sample names used for training
- Decisions have new columns:
sample.intrain
,sample.ref
,sample.ist
- Fixed strange bug in
NAMESPACE
(watch out for this in the future), wheredo.isx
would be exported as aS3
method calleddo
. Now roxygen generates the correct file again - Minor changes
- Important API change:
ref.y
does not exist anymore. Now one can control the samples included in the models usingid.oc
,id.bin
and those that are IST'd withid.ref
andid.ist
- TODO: make use of these in
plot()
and batch mode
reduce.reg()
can now wait for unfinished jobs and pass arguments toloadRegistry()
- The wrapper
do.isx()
does not return the models now, to save disk space and hopefully lower execution time inreduceResults()
- Updated the main
README.md
to give an overview of IST's purpose
- Fixed bug in
get.de.genes()
that would always restrict to 100 differential genes - Documented
pathways.table.oc
andpathways.table.bin
in wrapper
- Seems like the
0.4.10
version batch jobs completed successfully (isd, isa, iss) - Fixed example for
define()
- Batch mode seems to work for pathway models
Comments:
- If pathway models are going to be independent of the translator,
they can be sped up by just fitting them once and then running all the
translators on them. Might need another wrapper like
do.isx()
.
This version contains important changes on how the wrappers work.
I am trying to have a common interface to avoid duplicated code.
This is achieved by combining ellipsis and explicit arguments that
are passed on, and leveraging the common
define()
, fit()
and predict()
API.
- It is unclear whether batch model is working yet.
- Removed duplicated entries in docs
- Defined a way to create objects:
define()
. It is likenew()
but it can ignore mismatching arguments in ellipsis without throwing error. - Now
do.isx()
can run bothist.translator
andist.pathways
.
- Fixed unit testing (change in behaviour of
limma
helper) - Fixed duplicated documented arguments
- Fixed uses of
1:n
- Fixed tabs (multiple of 4)
- Removed lines of length >80
- Updated documentation format
- Added
biocViews
- Updated maintainer/author fields to be compliant
BiocCheck()
gives 1W (set.seed) and 5N
- Batch jobs can now run
isd
,isa
ansiss
- Changed name of wrapper to run the three of them
- Added imports from
stats
andutils
- Helper to find DE genes will only return their ids by default,
topTable
can be obtained with a flag - Batch can run with and without list of
de.genes
- Jackknifing is not run by default on pls models
- Added
statmod
package for robustlimma
estimates
- Added
limma
for CI/CD - Now
fit()
triggers the DE genes analysis - Added unit testing in workflow
- Fixed unit testing
- Added function to trigger
limma
analysis to findde.genes
- Moved unit testing of defaults to their own file
- Fix bug where 0-confidence orthologs would make it through
- Intra-species mapping now has same format as inter-species
- Cleaned up log files
- Small fix to skip batch tests
- Added the required packages
SummarizedExperiment
,batchtools
to CI/CD - Build ignores
reg-
andMeta
directories
This commit contains the very basics for batch analysis, focused on one flavour for now (ISD). This affects a good amount of files.
- Main function: wrapper
do.batch.isd()
- Default SGE configuration files
default.config.batch()
- Helper for reducing:
reduce.reg()
(UNTESTED) - Added batch configuration files for regular job and e-mail notification
(just works for my e-mail right now) in
inst/templates
- Basic examples and nit testing
and basic submitter wrapping
do.isd()
.gitignore
will ignore directories starting byreg-
Other changes
- Helpers now have a separate file from defaults
To do
- Check if batch version is actually using predefined orthologs
- Make sure no small files are left on the working directory (logs...)
- Test the reduce function
- Added
do.isd()
wrapper, documented and with example
fit.ist.translator()
should re-use orthology mapping if provided- Added test case in workflow to cover this
- Removed
mapOrthologs()
function, was apparently a duplicate oforthIDcon()
and was never called - Changed mistaken
utils::new()
tomethods::new()
- Added the option to specify a
proxy
in the ortholog mapperorthIDcon()
, checked the function arguments and added ellipsis
- Fixed
.gitlab-ci.yml
- Still dealing with vignette prebuilt index
- Info in github and in SO (link1 link2)
- Now I think this warning is because of the
--no-build-vignettes
option - Modified the
.gitlab-ci.yml
file because the error saidBiocStyle
was not available - Importing
utils
because now new objects are created withutils::new()
- Trying to fix the following error in builder:
Package has a VignetteBuilder field but no prebuilt vignette index.
- Changed
.Rbuildignore
to not ignore^doc$
and^Meta$
- Added quickstart vignette by taking the example from
data-raw/01_showcase.R
- Fixed small erorr in showcase
- Small changes in
.Rbuildignore
,.gitignore
- Now only 20 repeats in the binary model training (down from 100)
- Expanded example scripts to try batch jobs
- Expanded example data for testing purposes
- Fixed small bug for intra-species translators
- Added
show
method forist.pathways
- Added
predict
method forist.pathways
- Added
fit
forist.pathways
- Added pathways to the sample data
- Added
ist.pathways
class
- Small fix
- CI enabled
- The one-class model sets seeds as well
- Seeds are optional now (can be useful for parallel jobs)
- Added helper functions to generate default parameters for binary classifiers,
was particularly annoying to set up seeds in
caret::train()
- Removed internal data with defaults
- Changed
preProc
topreProcess
, it's unclear why this is referred to as both in the documentation (?train
) - Seeds are controlled in the binary classifier, two approaches:
with a
seed
argument, which calls toset.seed()
before training, and using theseeds
built-in argument intrainControl
. Still have to check how this works in a parallel setup.
- Changed the validation scheme for the binary classifier
plsr
- now usescaret
and runs by default 100 rounds of repeated CV- Extra slot with the
train
object - Supress only the warning on regression using two values only
- When fitting the final PLS, all the specified number of components
will be fit (especially for scoreplot purposes); however, when predicting
the decision, only the optimal
Ncomp
will be taken into account - By default, jackknifing is enabled on the final model
- Fixed bug in
predict.ist.bin()
that would not predict scores even if asked to do so - Added requirement on the checkmate package version
- Added
plot
method for scores and decision values - Doc and test updates
- Added the
ref.y
slot to iss to have traceability - Improved
show
- Added true fold changes to sample data
- Showcase is longer and has density plots
- Improved documentation on
fit
andpredict
methods - Added
show
methods
- Now
predict
can give the scores of thepls
models (isd/isa) - Defaults to 2 components
- Added example (future vignette)
- Refactored
type
toflavour
- Small class refactoring
- Fixed bug in
predict
- Removed unnecessary file for
predict
method
- Added
predict
method for the discriminator
- Removed
isc
, now it is a particular case ofisa
- Changes to the class definitions (more checks, slot removal)
- Primitive workflow already available
fit
method for the discriminator that applies to all flavours
- Refactoring of class and methods names (were too long)
- Added a
predict
method for binary classifier
- Added class for the classifier model in ISS
- Added a
fit
method for it
- Added a
predict
method for one-class
- Added class for the one-class model in ISS
- Added a
fit
method for it
- Same as version
0.1.7
with the predict method
- Now the fit method is in its own file and better commented
- Added predict method for the translator
- Modified the fit method, to join the mapping data.tables once only
- Updated the format of the sample data
- Updated unit testing
- Fixed fold change data frame (gene ids must be character)
- Added fit method for the translator
- Added intra-organism mapper
- Added classes for translator and discriminator
- Added sample data
- Fixed doc
- Modified
DESCRIPTION
- Added basic orthology mapper
- Initial addition of functions and tests