NEWS

---

GOSemSim 2.39.4

• let godata() take a 'GSON' object of GO annotations as annoDb, closing the ## to be supported branch of #44. The branch was already present but could never fire -- it tested inherits(annoDb, 'gson') while the S4 class is GSON -- so passing e.g. the output of clusterProfiler::gson_GO() fell through to check_goAnno() and failed with "annoDb as a data.frame should contains 'GO' and 'ONTOLOGY' columns". GSON carries no ontology information, so ONTOLOGY is recovered from 'GO.db' term by term; against AnnotationDbi::toTable(GOTERM) on an 80,000-row sample of gson_GO(org.Hs.eg.db, ont = "ALL") the resulting BP/CC/MF split (51058/16942/12000) agrees with GO.db for every term. GO IDs held without the GO: prefix are recognised, terms absent from 'GO.db' are dropped with a warning, and a gene set containing no GO IDs at all is now rejected instead of quietly yielding an empty dataset. 'gson' goes to Suggests rather than Imports, because reading the slots autoloads the namespace and the route should cost nothing to users who do not take it (2026-10-08, Thu)

GOSemSim 2.39.3

• take the set of terms to score in computeIC() from the installed 'GO.db' instead of from the gotbl table shipped in data/. That table was last rebuilt in 2021 and 524 BP terms added to GO.db since then are missing from it, so goSim() and geneSim() returned NA for them even though their descendants carry annotations and their IC is computable -- 167 of them are annotated in org.Hs.eg.db, plus 45 for CC and 367 for MF (#33). The missing terms also biased every other IC slightly, since they count towards the denominator but can never contribute to an ancestor's descendant sum; about 0.3% of terms were overestimated by more than 0.5 (2026-09-22, Tue)
• import a symbol from 'Rcpp', so that the compiled infoContentMethod_cpp() can actually be called. 'Rcpp' was listed in LinkingTo only, so loading GOSemSim never loaded the Rcpp namespace, and the Rcpp::RNGScope in the generated wrapper resolved enterRNGScope through R_GetCCallable("Rcpp", ...) only if the user happened to attach Rcpp first. Without it every Resnik, Lin, Jiang and Rel call failed with function 'enterRNGScope' not provided by package 'Rcpp' (2026-09-21, Mon)
• cache the ontology mappings behind getAncestors(), getParents() and getOffsprings(). Each call rebuilt its whole mapping -- AnnotationDbi::as.list(GOBPANCESTOR) and friends cost about 0.5 s -- while callers only ever pick out one or two terms, so the same mapping was rebuilt hundreds of times within a single geneSim() call. The mappings are now cached per ontology, which makes TCSS and the IC-based measures several times faster without changing any result (2026-09-21, Mon)
• fix TCSS similarity collapsing to a constant. A cluster is normalised by the largest finite IC among its members, so a cluster with exactly one finite-IC member always yielded an ICA of 1, however general that term was. GO:0008150, the root of BP, is such a cluster at the default cutoff and is a common ancestor of every BP pair, so max(sim_value) returned 1 for all 300 random BP pairs tested and mgeneSim(measure = "TCSS") was an all-ones matrix. The collapse also made the objective of tcss_cutoff() degenerate: at cutoffs of 3.5 and above the predictions were constant and AUC was pinned at exactly 0.5, so only F1 carried any signal. Such clusters are now normalised by the global maximum IC instead. The change is confined to exactly those clusters -- the set of clusters whose ICA changes equals the set with a single finite-IC member, at every cutoff and ontology tested -- and it removes the same collapse at other settings, e.g. GO:0050896, GO:0046683, GO:0014074, GO:0018193 (BP), GO:0016651, GO:0052745, GO:0003824 (MF) and GO:0005575, GO:0098644, GO:0019898 (CC) (2026-09-21, Mon)

GOSemSim 2.39.2

• fix termSim(..., method = "TCSS") for terms belonging to more than one cluster, where ica[[clusid[[term]]]] attempted recursive indexing ("subscript out of bounds") (2026-09-21, Mon)
• add set_auto_update() to control auto-update of ontology databases (2026-06-29, Mon)

GOSemSim 2.39.1

• improve pairwise similarity performance by reducing repeated matrix scans in score combination (2026-06-28, Sun)
• support optional BiocParallel parallel pairwise calculation in mgeneSim() and mclusterSim() (#31, 2026-06-28, Sun)
• improve TCSS robustness and document TCSS usage with topology cutoff preprocessing (2026-06-28, Sun)
• fix Wang method semantic value cache to avoid reusing GO term values across ontologies (#30, 2026-06-28, Sun)
• fix Wang method edge weighting for GO relationships by normalizing isa and part of relationship names (#51, 2026-06-28, Sun)

GOSemSim 2.38.0

• Bioconductor RELEASE_3_23 (2026-04-29, Wed)

GOSemSim 2.37.2

• refactor .initial() to use a controlled environment for robust data loading (2026-01-15, Thu)
• enhance cache reading robustness in computeIC() and wangMethod_internal() (2026-01-15, Thu)
• update load_onto() documentation (2026-01-12, Mon)
• enhance godata() input validation (2026-01-12, Mon)
• refactor internal ontology type checking (2026-01-12, Mon)
• fix cache integration in .initial and refactor consumers to use get_cache_element (2026-01-12, Mon)
• fix load_onto() using cache incorrectly (2026-01-12, Mon)
• use cache mechanisms defined in 'yulab.utils' for caching (2026-01-12, Mon)
• update URLs, fix typos and refine roxygen documentation (2026-01-12, Mon)

GOSemSim 2.37.1

• use cache mechanisms defined in 'yulab.utils' for caching (2025-12-08, Mon)
• update roxygen to use 'markdown' syntax and unify duplicated docs into templates (in 'man-roxygen' folder) (2025-12-07, Sun)
• move get_organism() from 'DOSE' (2025-12-07, Sun)
• move load_OrgDb() to 'yulab.utils' (2025-12-05, Fri)

GOSemSim 2.36.0

• Bioconductor RELEASE_3_22 (2025-11-01, Sat)

GOSemSim 2.35.2

• bug fixed that introduced in previous commit (2025-10-03, Fri)
  • https://github.com/YuLab-SMU/DOSE/issues/86#issuecomment-3361787836

GOSemSim 2.35.1

• alternative url for load_onto() (2025-08-12, Tue)

GOSemSim 2.34.0

• Bioconductor RELEASE_3_21 (2025-04-17, Thu)

GOSemSim 2.32.0

• Bioconductor RELEASE_3_20 (2024-10-30, Wed)

GOSemSim 2.31.1

• get_rel_df() to access ontology relation data frame required by the Wang method (2024-08-13, Tue)
• a set of 'OntDb' methods for accessing sqlite data generated by the 'obolite' package (2024-08-13, Tue)
• use yulab.utils::yulab_msg() for startup message (2024-07-26, Fri)

GOSemSim 2.30.0

• Bioconductor RELEASE_3_19 (2024-05-15, Wed)

GOSemSim 2.29.2

• update buildGOmap() parameter to be consistent with enricher() and GSEA() (2024-02-06, Tue, #47)

GOSemSim 2.29.1

• extend godata() to support passing a data.frame (can be output of read.gaf() or read.blast2go()) to 'annoDb' (2023-01-16, Tue)
• deprecate 'OrgDb' and introduce new parameter 'annoDb' in godata()
• standardize the output of read.gaf() and read.blast2go()
• optimize buildGOmap()

GOSemSim 2.28.0

• Bioconductor RELEASE_3_18 (2023-10-25, Wed)

GOSemSim 2.27.3

• use check_installed() to check package dependency (2023-09-12, Tue, #43)

GOSemSim 2.27.2

• read.blast2go() to parse 'blast2go' result (2023-07-10, Mon, #41, #42)
• move buildGOmap() and read.gaf() from 'clusterProfiler' (2023-07-10, Mon)

GOSemSim 2.27.1

• semantic similarity measurement support for MPO (2023-04-06, Thu)
• TCSS semantic similarity measurement support for DO and MPO (2023-04-06, Thu)

GOSemSim 2.26.0

• Bioconductor RELEASE_3_17 (2023-05-03, Wed)

GOSemSim 2.24.0

• Bioconductor RELEASE_3_16 (2022-11-02, Wed)

GOSemSim 2.23.1

• Replacing DO.db with HDO.db (2022-07-29, Mon)

GOSemSim 2.22.0

• Bioconductor 3.15 release

GOSemSim 2.21.1

• Avoid eval-parse in load_OrgDb() (2022-01-10, Mon)

GOSemSim 2.20

• Bioconductor 3.14 release

GOSemSim 2.19.1

• TCSS method (@qibaiqi, #35; 2021-08-02, Mon)

GOSemSim 2.18.0

• Bioconductor 3.13 release

GOSemSim 2.17.1

• bug fixed according to the update of GO.db (2020-10-29, Thu)
  • https://github.com/YuLab-SMU/GOSemSim/issues/32

GOSemSim 2.16.0

• Bioconductor 3.12 release (2020-10-28, Wed)

GOSemSim 2.15.2

• new site, https://yulab-smu.top/biomedical-knowledge-mining-book/ for documentation (2020-09-04, Fri)
• update vignette
• update data/gotbl

GOSemSim 2.15.1

• bug fixed of IC method when input IDs contain invalid terms. (2020-07-25, Sat)

GOSemSim 2.14.0

• Bioconductor 3.11 release

GOSemSim 2.13.1

• add new citation (2020-03-19, Thu)
• fixed compiling error due to the change of Rcpp
  • https://github.com/YuLab-SMU/GOSemSim/issues/27

GOSemSim 2.7.1

• mgeneSim and mclusterSim now always return matrix (2018-08-08, Wed)
  • https://www.biostars.org/p/330642/#331598

GOSemSim 2.5.1

• return NA for deprecated IDs (2018-01-09, Fri)
  • https://support.bioconductor.org/p/105822/#105840