Skip to contents

Changelog

MotrpacHumanPreSuspensionAnalysis 2.0.8

Changes

  • plot_enrich_heatmap() gains return_drawing: with TRUE it returns draw, width, height and n_sets instead of writing a PDF, and filename is optional. It also names any set_ids it drops.

Dependencies

  • MotrpacBicQC is required at >= 2.0.0, the v2.0.0 release tag (2026-09-23), and DESCRIPTION declares Remotes: MoTrPAC/MotrpacBicQC@v2.0.0 so pak resolves it from GitHub. 2.0.0 drops the inspectdf import, which is archived on CRAN and made MotrpacBicQC unresolvable from a clean library; assay_codes is unchanged from 1.9.0.

  • gridtext is added to Imports: plot_feature_heatmap() renders its column title with ComplexHeatmap::gt_render(), which needs it, and ComplexHeatmap only suggests it.

  • cmapR (Bioconductor) is added to Suggests. PTMSEA_INPUT is a list of cmapR::GCT S4 objects, and on a machine without cmapR R CMD check failed its data inspection with a WARNING (“unable to load required package ‘cmapR’”) because no field declared the package. The object still loads and its slots are reachable without cmapR; printing it or using the cmapR accessors needs it installed.

Internals

  • A GitHub Actions workflow, R-CMD-check.yaml, now runs R CMD check — tests and vignettes included — on every pull request and on pushes to main. Until now only the pkgdown site build ran, and only after a merge, so a failing test could reach main unseen.

  • The roxygen2 hook in R/zzz.R calls utils::getFromNamespace() with its namespace, clearing the “no visible global function definition for ‘getFromNamespace’” NOTE that R CMD check has reported since before 2.0.

  • R-CMD-check.yaml runs the check under a virtual display (Xvfb): Mfuzz loads Tk, which otherwise warns at install time on a headless runner.

Documentation

  • The README’s Data Use Agreement cites the c2.0 collection; it still named data release 1.3.0. The documentation now names the collections c1.3 and c2.0 throughout, matching the release paths, where it had written v1.3 and v2.0.

  • The versioning policy in the 2.0.0 entry below, and the README’s versioning note, now say what the version means: the first two digits name the data collection (every 2.0.x release ships c2.0) and the last digit counts package updates. The README also records that the 2.0.x series is the basis of the forthcoming pre-print, while 0.2.0 remains the version behind the initial one.

MotrpacHumanPreSuspensionAnalysis 2.0.7

New data

  • ORA_COLORS — the white-to-#543483 ramp for ORA heatmaps, as c(low, high); pass it to TMSig::enrichmap(colors = ). plot_cluster_enrichment() now reads it.

  • PTMSEA_RESULTS — PTM-SEA results for the prot-ph EE-CON and RE-CON contrasts (muscle 506 signatures x 6 contrasts, adipose 437 x 2), in the long layout of CAMERA_RESULTS with NES in place of t, df and z.std. Provenance in data-raw/PTMSEA/README.md.

Removed data

  • The vendored assay_codes object is removed, along with its man page and data-raw/assay_codes.R. It was a 44-row snapshot of MotrpacBicQC::assay_codes, taken when inspectdf was archived on CRAN and MotrpacBicQC could not be installed from a current snapshot. The snapshot has been retired in favour of reading upstream live. Code that referenced MotrpacHumanPreSuspensionAnalysis::assay_codes should read MotrpacBicQC::assay_codes.

Data objects

  • HUMAN_FEATURE_TO_GENE drops confident_site and has 12 columns where 2.0.3 gave it 13. The flag is measured per tissue, and this table is keyed on (assay, feature_id) with no tissue column, so the only value it could carry was the collapse across tissues — not the measurement for either tissue on the 859 of 7,865 shared prot-ph sites where muscle and adipose disagree. Read the per-tissue value from *_PROT_PH_QC$feature_metadata in MotrpacHumanPreSuspensionData, which is what preprocess_PTMSEA() does; PTMSEA_INPUT never read this table and is unaffected. All 1,920,618 rows and the other 11 columns are unchanged from 2.0.3.

Changes

  • plot_single_feature() reads MotrpacBicQC::assay_codes directly. Upstream 1.9.0 adds rows for metab-t-clinical, prot-clinical, metab-t-conv and metab-t-imm-crt, none of which existed in the vendored snapshot, so the hard-coded label fallback for the clinical omes is deleted — those facet strips are now labelled from upstream and read Clin. Chem rather than Clin. Chem.. The metab-t-conv label is still overridden to Conv. Metab (log2): upstream labels it Conv(T) in the LAB family, which does not distinguish it from its log2 twin metab-t-clinical.

  • Upstream also revises the 44 pre-existing rows in two columns this package does not display: assay_name punctuation on six rows (comma to hyphen) and cas_code on four (transcript-rna-seq and transcript-rna-seq-splicing from mssm to stanford, prot-pr and prot-ph from pnnl to broad_prot). Every assay_short_text is unchanged, so no existing figure label moves.

  • load_differential_analysis(epigen = TRUE) reads the c2.0 epigenomics DA from the public CloudFront release again, with no bucket access or local cache. load_differential_analysis() and plot_single_feature() drop the gsutil and bucket arguments; repo_local_dir is kept but ignored, with a message.

  • plot_enrich_heatmap() accepts PTMSEA_RESULTS again, plotting NES; set_ids takes PTMsigDB signature IDs for PTM-SEA input. n_top breaks p-value ties by the absolute statistic, so PTM-SEA’s permutation-floor p-values no longer pull in every tied set.

  • data-raw/google_cloud_bucket_checks/ is removed; the bucket validation pipeline lives in motrpac-human-presuspension-repro.

Documentation

  • The 2.0.0 entry below gains the collection-level provenance behind the c2.0 regeneration and the versioning policy that governs it. Those objects have shipped since 2.0.0; only the record of them is new.

MotrpacHumanPreSuspensionAnalysis 2.0.6

Changes

  • plot_feature_heatmap(): multi_tissue_clust_rows = TRUE works (it referenced an undefined object); new right_annotation, heatmap_args, draw_args and return_drawing arguments.

Data objects

  • COVARIATES_FILE drops the BMI and codedsiteid rows for epigen-methylcap-seq, which the MALAX model (~0 + group_timepoint + age + sex + (1 | pid)) never included.

MotrpacHumanPreSuspensionAnalysis 2.0.5

New data

  • PTMSEA_INPUT — the prot-ph differential-analysis z-statistics as a PTM-SEA input, one GCT per tissue, from the confidently localized sites. GCT is cmapR’s S4 class; attach cmapR to use its accessors.

Data objects

  • The thirteen *_DA objects gain CI.L_calculated and CI.R_calculated, the 95% confidence interval on logFC, computed against each contrast’s own residual degrees of freedom. No existing column or row changed. topTable’s CI.L/CI.R remain unshipped: for a dream fit they bound every contrast by the first contrast’s degrees of freedom.

  • UTORONTO_TFs is now the prot-ph TF regulator pool — 1,381 rows, feature_id and gene_symbol — where it was the raw UToronto extract, 2,765 rows and 28 columns keyed by Ensembl gene ID. The annotation columns are gone and a row is a phosphosite, not a gene.

MotrpacHumanPreSuspensionAnalysis 2.0.4

Data objects

  • FCM_CLUSTERS clusters blood into 12 clusters where it had 13; adipose (13) and muscle
    1. are unchanged and reproduce bit for bit. FCM_CAMERA and FCM_ORA follow the new clustering and have 434,337 rows where they had 444,806. Cluster numbers are a trajectory order, not identities, so blood cluster N in 2.0.4 is not blood cluster N in 2.0.3 — against the outgoing objects blood scores an adjusted Rand index of 0.711.

MotrpacHumanPreSuspensionAnalysis 2.0.3

Data objects

  • HUMAN_FEATURE_TO_GENE gains three columns and has 13 where it had 10. No row and no existing column changed: all 1,920,618 rows and all ten previously shipped columns are identical to 2.0.2, so nothing that reads the table today reads anything different. Code selecting columns by position has to be updated; code selecting by name does not.

    • custom_annotation (factor) and relationship_to_gene (numeric) say where an ATAC-seq or MethylCap-seq peak sits relative to the gene it was assigned to — the region it falls in ("Promoter (<=1kb)", "Intron", "Distal Intergenic", and seven others) and the signed base-pair distance to that gene, 0 where the peak overlaps it. Both were produced by the pipeline all along and dropped before the table was built, so an epigenomics feature arrived carrying only a gene: a peak in a promoter and a peak 40 kb into an intron were indistinguishable once mapped. They are populated for all 1,852,716 epigenomics rows and NA everywhere else.

      Unlike confident_site they are not a per-tissue measurement — they are derived from the peak coordinates in the feature_id — so they take the same value in every tissue a peak appears in and are not collapsed. This was checked rather than assumed: none of the 306,788 ATAC or 1,545,930 MethylCap feature identifiers shared between tissues disagree.

    • confident_site (logical) is the phosphosite localization flag, NA outside prot-ph. It is collapsed across tissues — TRUE only where a site is confidently localized in every tissue that measured it — because this table is keyed on (assay, feature_id) and has no tissue column, and 859 sites disagree between muscle and adipose. Read *_PROT_PH_QC$feature_metadata in MotrpacHumanPreSuspensionData when tissue-specific localization matters; preprocess_PTMSEA() already does, and is unaffected by this addition. Removed again in 2.0.7 for the reason given there: a collapse across tissues is not the measurement for either tissue.

Documentation

  • ?HUMAN_FEATURE_TO_GENE documented assay as a factor. It is a character vector, and has been for as long as the table has been built this way.

  • The documented assay values did not include "prot-clinical", which the 2.0 split of clinical chemistry into a metabolomics and a proteomics assay introduced. All eight values the column actually takes are now listed.

MotrpacHumanPreSuspensionAnalysis 2.0.2

Changes

  • load_differential_analysis(epigen = TRUE) reads the epigenomics tables from Google Cloud Storage via gsutil instead of the public CloudFront release, and defaults to the current motrpac-human-presuspension-repro staging bucket. It now requires repo_local_dir, and gains gsutil and bucket arguments; plot_single_feature() passes all three through.
  • The unexported load_DA_from_AWS() and .load_single_ome_tissue_AWS() are removed. Its pinned version = "1.2" no longer matched the atac-seq tables, which are at v2.0.

MotrpacHumanPreSuspensionAnalysis 2.0.1

Data objects

  • The *_SUM_STATS objects are named, ordered and keyed the way the *_DA objects are. Every metabolomics platform is now labelled assay = "metab" with the platform in its own platform column, where before the platform was written into assay and there was no platform column.

    There are 17 objects where there were 46. The research metabolomics platforms are no longer one object each: they are stacked into a single {TISSUE}_METAB_SUM_STATS per tissue

  • load_summary_stats() returns the research metabolomics platforms as a single "metab" element per tissue rather than one element per platform — the nesting load_differential_analysis() returns, so the two tiers can be walked together. Naming one platform still loads them all, as before.

Bug fixes

  • plot_single_feature() draws the same legend for every tissue, whether or not that tissue has a timepoint below the p threshold. Combining plots with patchwork::plot_layout(guides = "collect") previously produced a repeated p threshold legend, because collection only merges guides that are identical and a tissue with nothing significant contributed a one-key legend. A single plot with no significant timepoints now shows both p threshold keys rather than only adj p >=.

  • plot_single_feature() plots clinical chemistry only when a clinical ome is requested. It previously appended the clinical rows whenever the feature name matched an analyte, so selected_omes = "transcript-rna-seq" with "Glucose" returned a clinical chemistry plot. selected_omes now accepts the omes in clinical_ome_list() by name and "all" includes them, matching how load_differential_analysis() treats clinical chemistry.

    A request that names another ome no longer returns clinical chemistry alongside it, and "metab" no longer implies "metab-t-clinical". Analytes measured both clinically and on a research platform, such as Cortisol and Lactate, return only what was asked for.

  • plot_single_feature() loads every non-epigenetic tissue and ome once and filters afterwards, rather than assembling the request ome by ome. Clinical chemistry is no longer a special case appended after the load, and the differential analysis and the summary statistics are put in one vocabulary before either is filtered.

    plot_single_feature("VEGFA") works again. The default selected_omes = "all" was broken for every non-metab feature by filter(platform != "metab-t-conv"): platform is NA on non-metab rows and filter drops NA, so the filter deleted the whole non-metab payload and the feature was reported as absent from the data.

  • plot_single_feature() no longer excludes the metab-t-conv platform, which is now plotted and labelled Conv. Metab (log2). It has no assay_codes row, so without that fallback its facet strip read NA. Note that it is the metab-t-clinical measurement on a log2 scale, so Glucose, Glycerol, KET and NEFA in blood now return a panel from each.

MotrpacHumanPreSuspensionAnalysis 2.0.0

Data objects regenerated by the motrpac-human-presuspension-repro pipeline for the c2.0 data collection.

Versioning

The package version names the data collection it ships. The first two digits are the collection: every 2.0.x release carries the c2.0 collection. The last digit counts package updates made on top of it, and such a release may change nothing about the data at all. When describing which data an analysis used, cite the collection (c2.0); the full package version identifies the code.

Within the collection, versioning is per file: a file is bumped to v2.0 only where its content actually changed, so the c2.0 collection legitimately contains files carrying earlier version suffixes. See the MoTrPAC Knowledge Center for the release and versioning policy: https://motrpac-data.org/knowledge-center.

Why the c2.0 objects differ from c1.3

  • Sample misalignment in QC-norm batch correction, affecting transcriptomics and Olink. limma::removeBatchEffect() pairs covariate row i with matrix column i positionally. The covariate table was built with merge(), which returns rows sorted by vialLabel, and was passed against a matrix whose columns were in count-file order (transcriptomics) or pivot order (Olink). The two orders are not the same, so each sample was batch-corrected using another sample’s batch, site and plate assignment. Median per-feature correlation against c1.3 is 0.925 for transcriptomics (0.908-0.944 across the three tissues) and 0.900 for prot-ol. Every differential-analysis and summary-statistic object on those two omes moves with it: counted at each feature’s best post-exercise timepoint on the endurance exercise-vs-control contrasts, blood prot-ol goes from 103 to 146 significant features of 1,417 Olink targets.

  • Replicate averaging in muscle proteomics (prot-pr, prot-ph). Muscle samples measured twice are meant to be merged by averaging the pair and dropping the now-redundant column. For intra-site pairs the mean was written into the column that was about to be deleted, so the value that survived was the first measurement on its own rather than the mean of the two. Inter-site pairs were averaged correctly, which is what made this easy to miss. Batch correction and replicate handling also ran in a different order in prot-pr than in prot-ph, so the two omes were not processed identically.

  • Feature metadata rebuilt. Each ome’s metadata_features is now a self-contained feature-to-gene mapping that matches its QC-norm matrix exactly. This is why every *_QC object in MotrpacHumanPreSuspensionData differs from c1.3 while only six have a qc_norm matrix whose values differ.

  • ATAC differential analysis refit. The released DA tables were built against an earlier feature set than the QC-norm matrices they accompany; the two are reconciled here. Muscle ATAC moves from 1,584 to 1,521 significant features on the delta-delta contrasts — 87 gained, 149 lost, logFC correlating 0.992 with the release. Blood (t05-pbmc) has no significant peaks, as in v1.3.

  • Six ATAC samples removed as sample mix-ups. OUTLIERS has 160 rows where c1.3 had 154; the six added are epigen-atac-seq samples, four blood and two muscle, now excluded from the analysis. No row was dropped.

  • One muscle transcript-rna-seq sample restored. It was missing from metadata_samples and was therefore dropped at the modelling stage. Differences in the refit differential analysis are minimal.

  • Metabolite names follow a frozen RefMet snapshot with updated overrides. 78 metabolomics feature_ids in the *_METAB_DA objects are renamed from c1.3, so joins on c1.3 names will miss them. The curated overrides gain 13 HODE -> 13-HODE and 20-HETE entries that previously resolved to NA (adipose 13-HODE now enters the DA). Other renames: * suffixes dropped (FA 18:1;O* -> FA 18:1;O), NAGly -> NA-Gly, NAD+ -> NAD, DHA -> Docosahexaenoic acid, Edetic acid -> EDTA, 13-Oxo-ODE -> 13-OxoODE.

Reference and enrichment objects

  • CAMERA_RESULTS has 1,016,991 rows, 1,869 fewer than c1.3, the losses concentrated in blood prot-ol (-1,419), adipose prot-pr (-270) and muscle prot-pr (-189): the gene universe moved with the feature-metadata rebuild, and a set that no longer intersects it is not scored. The ranking is largely intact — Spearman 0.957 to 1.000 on the signed statistic.

  • METABOLOMICS_CVS has 4,000 rows where the c1.3 object had 3,878, and the 122 rows are mostly a c1.3 defect rather than a c2.0 change: the c1.3 .rda was staler than the c1.3 .txt it was meant to mirror, which already published 4,000 keys. The change that actually propagates downstream is 69 lowest_CV flips and 93 refmet_name corrections, which together decide which copy of a duplicated RefMet name survives de-duplication.

  • OME_TISSUE_CODE has 53 rows where it had 49: it gains the two clinical omes (metab-t-clinical, prot-clinical) and the five lab-* plasma-chemistry tiers, and loses metab-meta-reg in all three tissues. The lab-* codes are inputs to the clinical omes and are deliberately absent from ome_available_list(); do not pass them to a loader.

  • "metab-meta-reg" is gone from ome_available_list(), OME_TISSUE_CODE and HUMAN_OME_COLORS (31 entries where there were 32). No object was ever built under that name, so every accessor offered it as a choice that returned nothing. Passing it to a loader now errors at match.arg(), and HUMAN_OME_COLORS[["metab-meta-reg"]] returns NULL.

Split data

  • BLOOD_CLINICAL_CHEMISTRY_SUM_STATS — replaced by the c2.0 split into BLOOD_METAB_T_CLINICAL_SUM_STATS and BLOOD_PROT_CLINICAL_SUM_STATS.
  • BLOOD_EPIGEN_ATAC_SEQ_SUM_STATS — no blood ATAC feature is significant at adj_p_value < 0.05 this cycle, and epigenomics summary statistics carry significant features only, so no object is built.
  • CLIN_CHEMISTRY_DA — replaced by the c2.0 split into BLOOD_METAB_T_CLINICAL_DA and BLOOD_PROT_CLINICAL_DA.

New data

  • BLOOD_METAB_T_CLINICAL_DA, BLOOD_METAB_T_CLINICAL_SUM_STATS, BLOOD_PROT_CLINICAL_DA, BLOOD_PROT_CLINICAL_SUM_STATS

    Clinical chemistry, one assay in c1.3, is split into a metabolomics and a proteomics assay.

Changes to data objects

  • 11 objects drop the CI.L, CI.R columns; code that selects them will error: ADIPOSE_METAB_DA, ADIPOSE_PROT_PH_DA, ADIPOSE_PROT_PR_DA, ADIPOSE_TRNSCRPT_DA, BLOOD_METAB_DA, BLOOD_PROT_OL_DA, BLOOD_TRNSCRPT_DA, MUSCLE_METAB_DA, MUSCLE_PROT_PH_DA, MUSCLE_PROT_PR_DA, MUSCLE_TRNSCRPT_DA.

  • HUMAN_FEATURE_TO_GENE gains the flanking_sequence column.

MotrpacHumanPreSuspensionAnalysis 0.2.4

This release addresses installation failures on R 4.6 and removes the Remotes: field that was breaking dependency resolution.

Verified with clean installs on R 4.4 (Bioconductor 3.20), R 4.5 (3.22), and R 4.6 (3.23).

MotrpacHumanPreSuspensionAnalysis 0.2.3

User-facing functions

  • clinical chemistry is labeled clinical-chemistry instead of clinical_chemistry
  • plot_single_feature() now clarifies clinical chemistry features correctly; the y-axis label now clarifies that clinical chemistry features are shown on an absolute scale while all other features are shown as log2(normalized value).
  • Updated HUMAN_FEATURE_TO_GENE following changes described in 0.2.2., which now maps to the features in qc-norm properly.

Internals and package checks

  • Namespace hygiene updates - proper imports have been labeled throughout
  • Trimmed globalVariables(), consolidated @importFrom tags, added the grid.rect import, filled in the package Description, and added RColorBrewer, doParallel, parallel, and randomForest to Suggests.

Backend: data generation and release tooling (data-raw)

These changes only affect analysts with sample-level data from MotrpacHumanPreSuspensionData.

  • Added a Google Cloud bucket validation pipeline under data-raw/google_cloud_bucket_checks/. The numbered scripts snapshot the production GCS bucket, copy it to an isolated staging folder, diff staging against the local updated files, upload changed files (removing superseded versions), and validate the staging structure and values against the installed package. A run_validation_pipeline.R driver orchestrates the steps; see the directory README.md for setup.
  • Moved ATAC peak annotation into generate_atac_qc_norm.R (.annotate_atac_features() / pre_cawg_get_peak_annotations_hs()) and refined the gene-mapping logic. Also updated the methylcap, clinical, and metabolomics generation scripts.

MotrpacHumanPreSuspensionAnalysis 0.2.2

Backend: Feature metadata gene annotation (data-raw)

These changes will only make functional differences for analysts with sample level data from MotrpacHumanPreSuspensionData

  • Added gene-level annotation to feature_metadata outputs for all proteomics omes. Previously, feature_metadata for Prot-PR, Prot-PH, and Prot-OL contained only raw provenance columns (UniProt accessions, PTM identifiers, redundant IDs) with no standardized gene symbol or Ensembl mapping. Feature metadata files now include gene_symbol, ensembl_gene, and entrez_gene columns derived from a three-round BioMart lookup strategy.
  • Added .annotate_olink() to generate_prot_ol_qc_norm.R: resolves Olink protein metadata (UniProt accession from uniprot_entry) to gene symbols and Ensembl IDs. Lookup proceeds via UniProt → gene symbol → Entrez ID, with three progressive BioMart passes to maximize coverage.
  • Added .annotate_prot_pr() to generate_prot_pr_qc_norm.R: extends the per-tissue feature_metadata_output from PR@rdesc with gene annotations. Uses feature_id (= protein_id, a UniProt accession) taken directly from the file as the BioMart lookup key.
  • Added .annotate_prot_ph() to generate_prot_ph_qc_norm.R: same annotation for phosphoproteomics. Uses protein_id (not feature_id, which is the PTM site identifier) taken directly from the file as the UniProt lookup key, preserving the full PTM feature_id in the output.
  • Prot-PR and Prot-PH annotation now resolves UniProt accessions to genes via a single BioMart getBM lookup (mirroring the Olink .annotate_olink() implementation), using the UniProt accession read directly from the source files rather than a downloaded UniProt ID mapping table.

MotrpacHumanPreSuspensionAnalysis 0.2.1

  • Added clinical chemistry differential analysis using the same structure as the molecular differential analysis; see: CLIN_CHEMISTRY_DA

MotrpacHumanPreSuspensionAnalysis 0.2.0

Documentation and website

  • Added pkgdown GitHub Actions workflow for automated website deployment.
  • Added pkgdown configuration updates for GitHub Pages publishing.
  • Clarified consortium-only optional functionality in the README.

Dependency handling

  • Made MotrpacHumanPreSuspensionData an optional consortium-only runtime dependency instead of a public package dependency for CI/pkgdown resolution.

MotrpacHumanPreSuspensionAnalysis 0.1.0

Initial release

  • Public release of summary statistics and modeling outputs from the MoTrPAC human pre-COVID suspension phase.
  • Added functions for loading differential analysis and summary statistics.
  • Added clustering, enrichment, and visualization utilities.
  • Added vignettes and package website scaffolding.