Skip to content

Changelog

Source: NEWS.md

SomaDataIO (development version)

SomaDataIO 6.1.0 🥳 🍾

CRAN release: 2024-03-26

Lifting Code 🚀

  • Major restructure of lift_adat() functionality (@stufield, #81, #78)
    • lift_adat() now takes a bridge = argument, replacing the anno.tbl = argument. Lifting is now performed internally for a better (and safer) user experience, without the necessity of an external annotations (Excel) file.
    • the majority of this refactoring was internal and the user should not experience a major disruption to the API.
    • much improved lifting/bridging documentation (#82)
  • Added a new lifting and bridging vignette (@stufield, #77)
    • in addition to the improved lifting documentation this new vignette provides additional context, explanation, clear examples, and lifting guidance.

New Functions ✨

  • is_lifted() is new and returns a boolean according to whether the signal space (RFU) has been previously lifted

  • Lifting accessor function for Lin’s CCC values (#88)

    • getSomaScanLiftCCC() accesses the lifting correlations between SomaScan versions for each analyte
    • returns a tibble split by sample matrix (serum or plasma)

  • merge_clin() is newly exported (#80)

    • a thin wrapper that allows users to merge clinical variables to soma_adat objects easily
    • previously users had to either use the CLI merge tool or merge in clinical variables themselves with dplyr

  • Newly exported ADAT “get**” helpers (#83)

  • Newly exported functions that were previously internal only:

New Vignettes 🤓

  • The package README is now simplified (#35)
    • example analysis workflows are now split out into their own vignettes/articles and cross-linked in the README
  • Reorganization and expansion of statistical vignettes (#35, #47)
    • moved 3 existing statistical examples from README into their own vignettes
    • resulting in four new “Statistical Workflow” vignettes/articles:
      • Binary classification via logistic regression
      • Linear regression for continuous variables
      • Two-group comparison via t-test
      • Three-group analysis ANOVA
  • Added new general analysis workflow vignettes
    • articles for the pkgdown website have been built out
    • new articles on:
      • safely mapping values among variables
      • safely renaming a data frame
      • loading-and-wrangling
      • typical train and test data splits
      • beginning the FAQs and/or Coming Soon pages
  • Added a new vignette describing how to use the command-line interface merge tool (#45)
    • the new CLI merge tool used to add new clinical data to existing ADAT file

Updates and Improvements 🔨

  • collapseAdats() better combines HEADER information (#86)

    • certain information, e.g. PlateScale and Cal*, are better maintained in the final collapsed ADAT
    • other entries are combined by pasting into a single string
    • should result in less duplication of superfluous entries and retention of more “useful” HEADER information in the resulting (collapsed) soma_adat

  • Update read_annotations() with 11k content (#85)

  • Update transform() and scaleAnalytes()

    • scaleAnalytes() (internal) now skips missing references and is much more like a “step” in the recipes package
    • transform() gets edge case protection with drop = FALSE in case a single-analyte soma_adat is scaled.

  • New row.names() S3 method support for soma_adat class

    • dispatched on calls to rownmaes()
    • rather than calling NextMethod() which normally would invoke data.frame, we now force the data.frame method in case there are tbl_df or grouped_df classes present that would be dispatched. Those are bypassed in favor of the data.frame because tbl_df 1) can nuke the attributes, 2) triggers a warning about adding rownames to a tibble.

  • New grouped_df S3 print support for the grouped soma_adat

    • now displays Grouping information from a call to the S3 print method for soma_adat class

  • New grouped_df S3 method support for soma_adat class (#66)

    • grouped_df data objects previously unsupported and were interfering with downstream S3 methods for dplyr verbs once NextMethod() was called
    • this support now ensures that the group methods are maintained, as well as the soma_adat class itself (and most importantly, with its attributes intact)

  • tidyr::separate.soma_adat() S3 method was simplified (#72)

    • now uses %||% helper internally
    • expanded error messages inside stopifnot() to be more informative

  • is_intact_attr() is now much quieter, signaling only when called indirectly (#71)

    • new conditional logic to silences signaling messages when called from within another function (indirectly)
    • these previously lead to confusing messages when they appear in wrappers, where is_intact_attr() can be, sometimes deeply, nested in the call stack

  • Development and improvements to the pkgdown website

    • added new links and improved clarity in YAML
    • added new logo at footer
    • restyled side bar for easier hyperlinking and getting help
    • clicking on the SomaLogic logo in the GitHub README now links to the pkgdown website
    • new “Coming Soon” drop-down section in the website header to let users know about active progress (but not yet ready for external publication)

  • SomaDataIO no longer depends on desc package

    • to generate the README.md

Internal 🚧

  • Internal rowname helpers were upgraded
    • they now use internal cross-functions as originally intended to avoid redundancy, efficiency, and improved debugging
  • sysdata.rda no longer contains non-exported functions (#59)
    • new internal helper functions:
      • convertColMeta()
      • genRowNames()
      • parseCheck()
      • syncColMeta()
      • scaleAnalytes()
  • Bug-fix for corner-case writing a single-analyte ADAT (#51)
    • RFU values are rounded to 1 decimal place when written by write_adat(), via a call to apply(), which expects a 2-dim object when replacing those values.
    • write_adat() no longer uses apply() and instead converts the entire RFU data frame to a matrix (maintains original dimensions), and use vectorized format conversion via sprintf()
    • in theory this should be faster because sprintf() is only called once on a long vector, rather than 1000s of times on shorter vectors (inside apply()).
  • Fixed missing closing parenthesis in SomaScanObjects.R (thanks @Hijinx725!, #40)

SomaDataIO 6.0.0 🎉

CRAN release: 2023-03-15

  • We are now on CRAN! 🥳

New changes

  • New clinical data merge CLI tool (@stufield, #25)
    • Rscript --vanilla merge_clin.R for merging clinical variables into existing *.adat SomaScan data files
    • added 2 new example meta.csv and meta2.csv files to run examples with random data but with valid index keys
    • see dir(system.file("cli", "merge", package = "SomaDataIO"))
  • Package data objects (@stufield, #32)
    • example_data.adat was reduced in size to n = 10 samples (from 192) to conform to CRAN size requirements (< 5MB)
    • the current file was renamed example_data10.adat to reflect this change
    • this likely has far-reaching consequences for users who access this flat file via system.file()
    • the example_data object itself however remains true to its original file (https://github.com/SomaLogic/SomaLogic-Data/blob/master/example_data.adat)
    • the directory location inst/example/ was renamed inst/extdata/ to conform to CRAN package standard naming conventions
    • the file single_sample.adat was removed from package data as it is now redundant (however still used in unit testing)
    • SomaDataObjects was renamed and is now SomaScanObjects
  • Gradual deprecation (@stufield)
  • New S3 print method default (@stufield)
    • tibble has new max_extra_cols = argument, which is set to 6 for the print.soma_adat method
  • New S3 merge method (@stufield, #31)
    • calling base::merge() on a soma_adat is strongly discouraged
    • we now redirect users to use dplyr::*_join() alternatives which are designed to preserve soma_adat attributes
  • Code hardening for prepHeaderMeta() (@stufield)
    • some ADATs do not have CreatedDate and CreatedBy in the HEADER entry. This currently breaks the writer
    • simplified to make more robust but also refactor to be more convenient (for abnormal ADATs not generated by standard SomaScan processing)
    • CreatedDateHistory was removed as an entry from written ADATs
    • CreatedByHistory was combined and dated for written ADATs
    • NULL behavior remains if keys are missing
    • CreatedBy and CreatedDate will be generated either as new entries or over-written as appropriate
  • Numerous non-user-facing (API) changes internal package maintenance, efficiency, and structural upgrades were included

SomaDataIO 5.3.1

  • Bug-fix release related to write_adat():

    • fixed bug in write_adat() that resulted from adding/removing clinical (non-SomaScan) variables to an ADAT. Export via write_adat() resulted in a broken ADAT file (@stufield, #18)
    • write_adat() now has much higher fidelity to original text file (*.adat) in full-cycle read-write-read operations; particularly in presence of bangs (!) in the Header section and in floating point decimals in the ?Col.Meta section
    • write_adat() no longer converts commas (,) to semi-colons (;) in the ?Col.Meta block (originally introduced to avoid cell alignment issues in *.csv formats)
    • write_adat() no longer concatenates written ADATs, when writing to the same file. Data is over-written to file to avoid mangled ADATs resulting from re-writing to the same connection and to match the default behavior of write.table(), write.csv(), etc.

  • read_adat() now has more consistent character type the Barcode2 variable in standard ADATs, now forces character class, does not allow R’s read.delim() to “guess” the type

  • Decreased dependency of magrittr pipes (%>%) in favor of the native R pipe (|>). As a result the package now depends on R >= 4.1.0

    • SomaDataIO will continue to re-export magrittr pipes for backward compatibility, but this should not be considered permanent. Please code accordingly

  • Migration to the default branch in GitHub from master -> main (@stufield, #19)

  • Numerous non-user-facing (API) changes internal package maintenance, efficiency, and structural upgrades were included

SomaDataIO 5.3.0

  • Upgrades primarily from improvements to SomaLogic internal code base, including: (@stufield)

    • general reduction on external package dependency to improve code stability
    • internal usage of base R alternatives to the readr package for parsing and importing ADATs (e.g. read.delim() over readr::read_delim()). This is mostly for code simplification, but can often result in marked speed improvements. As the SomaScan plex size increases, this speed improvement will become more important.
    • parseHeader() was dramatically simplified, now reading in lines 20L at a time until the RFU block is reached. In addition, once the block is reached, all header lines are read-in once and indexed (as opposed to line-by-line).
    • read_adat() now specifies column types via colClasses = which for the majority of the ADAT is type double for the RFU columns. This should dramatically improve speed of ingest.
    • write_adat() was simplified internally, with fewer nested apply and for-loops.
    • encoding for all input/output (I/O) is assumed to be UTF-8.

  • New getAnalytes() S3 method for class recipe from the recipes package.

  • New loadAdatsAsList() to load multiple ADAT files in a single call and optionally collapse them into a single data frame (@stufield, #8).

  • New getTargetNames() function to map ADAT seq.XXXX.XX names to corresponding protein targets from the annotations table

SomaDataIO 5.2.0

  • SomaLogic Inc. is now SomaLogic Operating Co. Inc.

  • Added new documentation regarding Col.Meta (@stufield, #12).

    • documentation around column meta data, row meta data, where they are found in an ADAT, and how to access them.

  • Research Use Only (“RUO”) language was added to the README (@stufield, #10).

  • Numerous internal code improvements from SomaLogic code-base (@stufield)

    • the consisted of reducing usage of external dependencies, e.g. using stop() over ui_stop() and warning() over ui_warn(), using usethis, cli, and crayon shims aliases.
    • package uses purrr very selectively and no longer uses stringr.
    • using base R alternatives in favor of increased stability for underlying, non-user-facing code.

  • New lift_adat() was added to provided ‘lifting’ functionality (@stufield, #11)

    • provides mechanism to convert RFU space between SomaScan versions (e.g. v4.1 -> v4.0).
    • added new S3 transform.soma_adat() method which simplifies linear scaling of soma_adat columns (analytes).
    • uses an “Annotations file” (Excel) as source of scalars for transformation.

  • Minor improvements and updates to the README.Rmd (@stufield, #7)

    • fixed a broken adat2eSet() link in README (#5).
    • clearer text to the README regarding Biobase installation.
    • added new links to external Bioconductor website in installation section of README.
    • new pkgdown and links to Issues (#4).
    • SomaLogic logo was added to README.
    • a lifecycle (“maturing”) badge was added.

  • Startup message was improved with dynamic width (@stufield).

  • New locateSeqId() function to pull out SeqId regex. (@stufield).

  • New read_annotations() function (@stufield, #2)

    • new function to parse/import SomaLogic annotations files (*.xlsx).

SomaDataIO 5.1.0

SomaDataIO 5.0.0

  • Initial public release to GitHub!