NEWS.md

In bmass: Bayesian Multivariate Analysis of Summary Statistics

Version News and Updates

Version updates will be tracked and explained here. Major updates & releases will be particularly highlighted.

bmass v1.0.3

Summary

• Patches for 2nd CRAN resubmission

Since previous version (v1.0.2)

Notes

• Changed \dontrun{} to \donttest{} in .Rd files

Next steps (if applicable)

• See v.1.0.0 'Next steps'

bmass v1.0.2

Summary

• Patches for CRAN resubmission

Since previous version (v1.0.1)

Notes

• Added some requested reformatting in DESCRIPTION

Next steps (if applicable)

• See v.1.0.0 'Next steps'

bmass v1.0.1

Summary

• Patches in-between first release and final CRAN submission/acceptance

Since previous version (v1.0.0)

Notes

• Fixed 'Vignette Title' entries for both vignettes
• Added in temp fix for R.3.6+ unit test re: use of RNG/sample()

Next steps (if applicable)

• See v.1.0.0 'Next steps'

bmass v1.0.0

Summary

• Finished prep for CRAN submission
• Setup Travis CI for repo
• Finished leftover unit tests
• Moved vignette setup into two introductory examples
• Edits and suggestions from Peter Carbonetto to help with first main release

Since previous version (v0.1.3)

• Finished doing necessary prep and checks for CRAN package submission
• Finished documenting leftover objects and function arguments for Roxgyen2 & devtools::check()
• Setup and connected Travis CI to git repo
• Finished unit tests in tests/testthat/test-BasicRFunctionality.R, test-ResultsFollowupAnalysisAndPlotting.R, test-Stephens2013PLoSONE.globallipids.GLCfuncs.R, and test-Stephens2013PLoSONE.test.funcs.R
• Made Models output more user-friendly versions
• Put phenotype names on top of Models and directed users in main vignette to utilize GetModelPriorMatrix() for a more interpretable version of ModelPriors
• Made a new introductory vignette using a small, simulated dataset, and moved the previous one into a 'more advanced' introductory example (referened as a 'real data' example)
• Merged and adopted edits made by Peter Carbonetto, as well as adopted a few of his suggestions
• Added proper releases on the github repo#
• Created a pkgdown website for the github repo
• Created the new vignette as described above
• Updated README.md
• Further cleaned up roxygen2 comments & examples

Notes

• Changed inputs of GetModelPriorMatrix() in ResultsFollowupAnalysisAndPlotting.R so that SigmaAlphas is given a default value to begin with and therefore positioned after LogFile

Next steps (if applicable)

• Change output of Matthew's code to use logBFs vs. lbf, and align with overall style of bmass wherever else applicable
• Create second vignette on how to run bmass() with some data already created (ie a merged dataset, a phenotype correlation matrix, and priors)
• Include note that, if provided, ZScoresCorMatrix should have the same phenotype order as DataSources
• Double-check new PassChecksForDataSources.R additions
• Flesh out CheckMergedDataSources as an analogue to CheckIndividualDataSources for when a merged dataset is provided.
• Include check that all datasets have same A1 (in MergeDataGWASAnnotateAndGetMarginalSNPs.R) & that A1 part of ExpectedColumnNames
• Include some of the extra, downstream functions such as GetModelPriorMatrix() in the advanced, introductory vignette

bmass v0.1.3

Summary

• Finished cleaning up all main comments leftover in every *.R file
• Added in Roxygen2 comments and the first vignette, and finished creating unit tests
• Went through devtools::check() results and made required edits

Since previous version (v0.1.2)

• Cleaned comments & formatting from bmass.R, Stephens2013PLoSONE.globallipids.GLCfuncs.R, & Stephens2013PLoSONE.test.funcs.R
• In bmass.R changed GWASThreshFlag from 0/1 to FALSE/TRUE
• Mmade related changes to DetermineAndApplyPriors in GetResultsFromMarginalSNPsAndFormat.R
• Switched GWASThreshFlag to default as TRUE
• In ResultsFollowupAnalysisAndPlotting.R removed a large number of the functions that were potentially more manuscript-specific and likely to become deprecated (most of which were plotting-related)
• Moved any library() calls to proper use of Imports in DESCRIPTION file

Notes

• Dealt with CHECK_0's in GetResultsFromMarginalSNPsAndFormat.R (ie CHECK_0 -- Prob: Throw an error/fail thing here)
• In bmass.R changed PrintLogStatements to PrintProgress
• In bmass.R removed the loadings of library(ggplot2) and library(reshape2)
• In bmass.R removed the commented out ExploreBestModelsUsingPosteriors()
• In bmass.R & main bmass() call added "A1" to ExpectedColumnNames

Next steps (if applicable)

• Finish doing necessary prep and checks for CRAN package submission
• Setup and connect Travis CI to git repo
• Change output of Matthew's code to use logBFs vs. lbf, and align with overall style of bmass wherever else applicable
• Fix Models and ModelPriors outputs to more user-friendly versions
• Put phenotype names on top of Models and possibly condense ModelPriors across all the SigmaAlphas
• Create second vignette on how to run bmass() with some data already created (ie a merged dataset, a phenotype correlation matrix, and priors)
• Include note that, if provided, ZScoresCorMatrix should have the same phenotype order as DataSources
• Finish unit tests in test-ResultsFollowupAnalysisAndPlotting.R, particularly evaluate issues with GetTopModelsPerSNPViaPosteriors
• Finish unit tests in test-Stephens2013PLoSONE.globallipids.GLCfuncs.R and test-Stephens2013PLoSONE.test.funcs.R
• Finish merge() unit tests in tests/testthat/test-BasicRFunctionality.R
• Flesh out CheckMergedDataSources as an analogue to CheckIndividualDataSources for when a merged dataset is provided.
• Double-check new PassChecksForDataSources.R additions
• Put in new checks for Marker column, input MergeDataSources file, etc... as well
• Include check that all datasets have same A1 (in MergeDataGWASAnnotateAndGetMarginalSNPs.R) & that A1 part of ExpectedColumnNames
• Check/correct how write(...stderr()) is functioning in bmass.R (since it seems like that might not be occuring like I was expecting it to?)

bmass v0.1.2

Summary

• Revised some important portions of data-preprocessing
• Began creating and flushing out ResultsFollowupAnalysisAndPlotting.R (eg GetModelPriorMatrix, GetTopModelsPerSNPViaPosteriors, ``, etc...)
• Began cleaning up and adding in proper comments/formatting/etc for eventual, formal, first release (ie v1.0.0)
• Followed up on a number of previous suggested comments/edits from prior work
• Added in GWASThreshFlag & GWASThreshValue functionality

Since previous version (v0.1.1)

• Separated MergeDataSources() from AnnotateMergedDataWithGWASSNPs() in main bmass() call since the two parts take very different times, ie former is longer than the latter.
• Changed AnnotateDataWithGWASSNPs to a quicker version where the loop is over the GWAS SNPs and not the merged dataset; taking advantage of in-line conditional statements as well for row selection.
• Included new flag PrintMergedData to indicate whether bmassOutput$MergedDataSources should be included as part of the output and not made NULL. This would be for whether a user wants to save a copy of MergedDataSources and reuse it for later runs.
• Added bmass() input variable ZScoresCorMatrix that allows users to manually use a specified phenotype correlation matrix.
• Fixed DetermineAndApplyPriors to output PreviousSNPs$logBFs in the sections when ProvidedPriors is not NULL and when UseFlatPriors is TRUE.
• Added in Marker column as an expectation for input files in lieu of the ChrBP column I constructed. If ChrBP is desired to be used as marker names, it can be created by users before submitting the files to bmass().
• Added in functionality to designate GWASsnps not only by input GWASsnps file but also by a GWAS p-value threshold (via GWASThreshFlag & GWASThreshValue)
• Properly designated PreviousSNPs sections in GetResultsFromMarginalSNPsAndFormat.R (both DetermineAndApplyPriors and FinalizeAndFormatResults) with ...if (!is.null(GWASsnps)) {... encapsulation as needed
• Added in write(..., stderr()) statements to mimic calls to LogFile1 to be produced as bmass() runs, eg giving user a live look into how/where the script is proceeding. These statements are a bit more broad/higher-level than the LogFile1 calls themselves. Also created PrintLogStatements (default FALSE) to turn this functionality on/off.
• In MergeDataSources, added in section post-merge that checks SNPs still have correct MAF (eg <= .5).
• Also now checking and dropping SNPs that appear 'fixed' in dataset, eg MAF is equal to 0.
• Moved MAF checks (eg if MAF <= .5, if fixed (0 or 1), etc...) to CheckIndividualDataSources section.
• Added in CheckDataSourceMAFIsMAF and CheckDataSourceMAFFixed functions.
• Additionally, now just quitting program if there are fixed SNPs or SNPs that are still > .5; rather have the user fix their datasets and make sure things are correct rather than doing processing mid-package/functions

Notes

• Moved bmass.devlog.vs1.txt to bmass.devlog.vs1.md.
• Changed adding max * length to just add max in GetSumAcrossSigmaAlphas_withPriors.
• Moved to data.frame column calls via $ from atomic vector style calling in AnnotateDataWithGWASSNPs after move away from apply function.
• Changed logBF to logBFs.
• Fixed check for ProvidedPriors in CheckIndividualDataSources to expect 3 ^ (length(DataSources) * length(SigmaAlphas)), not 3 ^ length(DataSources).
• In bmass.R changed if (is.null(PrintMergedData)) { to if (PrintMergedData == FALSE) {, and changed default value of PrintMergedData from NULL to FALSE.
• Changed posteriorprob so that prior in apply(prior * 10^lbf,2,normalize) now specifies the proper matrix dimensions rather than assuming, given proper length, prior will be multiplied along lbf filling column-wise first (which it does by default).
• Added MergedDataSources as one of the return values of ProcessMergedAndAnnotatedDataSources in MergeDataGWASAnnotateAndGetMarginalSNPs.R
• Added SigmaAlphas as an input to GetModelPriorMatrix in ResultsFollowupAnalysisAndPlotting.R
• Corrected ZScoreHitFlag1 cutoff by including 2*pnorm...
• Added if (nrow(SummaryOfTopModels) > 1) {... to GetTopModelsPerSNPViaPosteriors in ResultsFollowupAnalysisAndPlotting.R because (apparently) reordering a single-row matrix in the way that I'm doing it messes up its classification as a matrix; after this command, and loss of matrix class, the following colnames() command throws an error. This does not occur when there are more than one row and an actualy 're-ordering' occurs (presumably).

Next steps (if applicable)

• Flesh out CheckMergedDataSources as an analogue to CheckIndividualDataSources for when a merged dataset is provided.
• Double-check new PassChecksForDataSources.R additions
• Put in new checks for Marker column, input MergeDataSources file, etc... as well
• Finish cleaning up comments/edits from offline/pre-dissertation work
• Start adding in necessary portions for proper R-package structure/eventual CRAN-submission (eg fleshed-out Roxygen2 comments, unit tests, vignettes, necessary additional datafiles for sections such as the unit tests and vignettes, etc...)
• Include check that all datasets have same A1 & that A1 part of ExpectedColumnNames
• Move towards getting v1.0.0 prepared

bmass v0.1.1

Summary

• First revision and reorganization/restructuring of .R files and internal code
• First steps in setting up unit tests via testthat

Since previous version (v0.1.0)

• Began moving chunks of code from PrepareData.R to better designated .R files
• Changed the few leftover instances of NewHits to NewSNPs
• Moved NewSNPs creation section to after creation of logBFs and PosteriorProbabilities matrices from MarginalHits; moved NewHits <- NULL initialization section of entire code
• Moved from LogFile1 to LogFile
• Changed ProvidedPriors to be the first option in the if/elseif/else block that also checks is.null(GWASsnps)..., etc...
• Began setting up unit tests via testthat, including for basic R functionality, PassChecksForDataSources.R, and
• Introduced main bmass input variable GWASsnps_AnnotateWindow
• Moved all instances of MarginalHits to MarginalSNPs, including in variable names, eg PruneMarginalHits and PruneMarginalHits_bpWindow
• Split up GetLogBFsFromData into GetLogBFsFromData and DetermineAndApplyPriors
• Moved GetResultsFromMarginalSNPs.R to GetResultsFromMarginalSNPsAndFormat.R, moved FinalizeAndFormatResults() back to GetResultsFromMarginalSNPsAndFormat.R, and removed FormatAndPrepareResults.R
• Created ExploreBestModelsUsingPosteriors to put code chunks dealing with posteriors and related analyses such as BestClass
• Moved FollowupResultsAnalysisFormattingAndPlotting.R to ResultsFollowupAnalysisAndPlotting.R
• Moved ExploreBestModelsUsingPosteriors into ResultsFollowupAnalysisAndPlotting.R

Notes

• Moved TestData1.txt, TestData2.txt, and TestSigSNPs.txt to bmass_TestData1.txt, bmass_TestData2.txt, and bmass_TestSigSNPs.txt
• Currently unclear whether Roxgyen-based .R files describing above test datasets should be in /data or in /R. R CMD check throws warning/error if .R files not in /data, but R won't load datasets via data() properly if .R files not found in /R
• Changed CheckCharacterFormat to CheckCharacterClass
• Changed underlying code for CheckDataSourceDirectionColumn, moving from a for loop to use of ...if (length(...[...!= "+" & ... != "-"]) > 0 )...
• Changed LogFile to bmassOutput$LogFile in main bmass.R file
• Using bmassOutput as a way to hold temp variables since can output and assign multiple variables via the list structure and self-made functions. Eg cannot do c(var1, var2) <- function(input1), but can do list1[c("var1", "var2")] <- function(input1) assuming function outputs a list with two entries
• Reworked order of input variables for main bmass function
• Included paste(..., collapse=" ") at end of stop() calls that had DataSources[!DataSources...]-type error messages
• stop() seems to already have some automatic paste()-type calls? Eg how it can incorporate my call to paste() at the end without any additional specifications
• Moved LogBF to logBF in all instances except for function name GetLogBFsFromData
• Moved if (!is.null(bmassSeedValue)) {... section to within the else{...} portion of the if (!is.null(ProvidedPriors)) {... block
• Changed Priors_Used to ModelPriors_Used
• In GetSumAcrossSigmaAlphas_withPriors changed ModelPriors to ModelPriors_Matrix
• In function call for GetSumAcrossSigmaAlphas_withPriors in FinalizeAndFormatResults, including ...matrix(...nrow=length(ModelPriors)...)... to make the number of rows explicit

Next steps (if applicable)

• Finish creating unit tests for all .R files and necessary functions
• Finish Roxygen commenting
• Continue revising/reviewing code, including working through remaining CHECK_0 flags
• Go through R CMD CHECK output?

bmass v0.1.0

Summary

• Reworked format of bmassOutput to include the list() structure of MaringalSNPs, NewSNPs, and PreviousSNPs

Since previous version (v0.0.4)

• Added the following to bmassOutput: NewSNPs, NewSNPs$SNPs, NewSNPs$logBFs, NewSNPs$Posteriors
• These are mostly focused on including the GWAS snp results
• Initialized bmassOutput variables and associated lists at beginning of function
• Changed SNPMarginalpValThreshold to SNPMarginalUnivariateThreshold and SNPMarginalMultivariateThreshold
• Moved PosteriorProbabilities to Posteriors
• Created MarginalSNPs, NewSNPs, and PreviousSNPs setup with list() formatting
• Seeing redundancy among MarginalSNPs, NewSNPs, and Previous SNPs, lead to this suggestion and change
• Moved back to PreviousSNPs from GWAShits
• Added BestModel_Posterior to PreviousSNPs mostly has a reminder/placeholder for the time being to expand on it later for both NewSNPs and MarginalSNPs

Notes

• Decided the creation of the first, fully working bmassOutput warranted moving to version 0.1.0
• Partially thinking should have moved to 0.1.0 from 0.0.3 but turns out there were a few useful edits to make to warrant 0.0.4 in the first place
• Next string of updates and 'small patches' will be geared towards reworking and cleaning up the organization of the R files and associated functions
• Then 0.2.0 on might focuse on creating useful/necessary follow-up analysis functions?
• Include BestModel as a variable for every SNP? Include posterior as well as the model itself?

Next steps (if applicable)

• Same as 0.0.4
• For follow-up functions: Put ModelMatrix, AllBut1Assoc, and MarginalPosteriors into this area?

bmass v0.0.4

Summary

• Created first draft of preliminary, complete bmassOutput variable

Since previous version (v0.0.3)

• Fleshed out first is.null(GWASsnps) & else subsections
• Implemented pruning section and PruneMarginalHits flag
• Created logBFs and PosteriorProbabilities output matrices
• Initialized GetSumAcrossSigmaAlphas_withPriors()
• Created NewSNPs output variable, in conjunction with GWASlogBFMinThreshold
• Added current desired, associated variables to bmassOutput

Notes

• Current content associated with bmassOutput:
• MarginalSNPs
• ModelPriors
• logBFs
• PosteriorProbabilities
• GWASlogBFMinThreshold
• NewSNPs

Next steps (if applicable)

• Next steps include:
• cleaning up code (eg superfluous comments, print statements, etc...)
• moving to a return() ending for the code via bmassOutput
• beginning to map out breaking up code into more definable subsections for better organized .R file subsets
• reviewing, updating, and changing (if/as necessary), bmassOutput associated variables
  • any other additions? include both logBFs and PosteriorProbabilities?
• begin considering downstream helper functions for follow-up analyses/plots of bmassOutput results

bmass documentation built on May 17, 2019, 9:02 a.m.