Review 1 of “A Multi-Source Profile of CALHM5: calcium homeostasis modulator family member 5 as an IDG Understudied Ion Channel” (v1)

Summary and recommendation

Recommendation: major revision. This review concerns version 1.

The manuscript compiles CALHM5 annotations from nine resources, reporting ARCHS4 coexpression partners, a tissue/cell-type expression summary, sparse GeneRIF coverage, one GO biological-process term, and 1,302 Harmonizome associations across ten datasets. A reproducible resource inventory could be useful, and the distinction between Enrichr gene-set membership lookup and enrichment testing is appropriate. However, the discussion applies an orphan-GPCR framework to an ion-channel-family protein, overlooks relevant CALHM5 research, interprets undefined similarity scores, and does not adequately document score provenance or missing-data handling. These issues affect the biological conclusions rather than only presentation.

I read the full version-1 Markdown and checked primary literature and resource documentation. I did not rerun the pipeline, inspect the linked images directly, or verify the attached PDF, which was inaccessible through my retrieval tools. Caption and table observations below concern the Markdown.

Major concerns and numbered requests

1. Replace the GPCR interpretation with a CALHM5-specific account. The title and classification identify an ion-channel target, whereas the therapeutic discussion calls CALHM5 an orphan GPCR and recommends endogenous-ligand discovery and G-protein coupling assays. NCBI Gene 254228 identifies CALHM5/FAM26E as a protein-coding calcium homeostasis modulator family member. The CALHM5-specific primary study “Cryo-EM structures of human calcium homeostasis modulator 5” reports human CALHM5 structures in 2020. Correct the GPCR claims throughout, including the metabolomics paragraph and coexpression interpretation. Replace the assertion that structural work awaits a future structure with a discussion of existing structures and their limitations. Frame future functional experiments around channel properties, localization, gating, and physiological context, distinguishing family-derived predictions from direct CALHM5 evidence.

2. Update the literature baseline and disease context. Two records in a queried GeneRIF dataset do not establish that only two PubMed papers concern CALHM5. Provide those PMIDs, search dates, source release, and alias handling, and label the metric as GeneRIF coverage. Add a targeted literature search using CALHM5 and FAM26E. In particular, Yang et al., 2026, DOI 10.1073/pnas.2509778123 report CALHM5 channel characterization and smooth-muscle calcium regulation, human aneurysm-tissue expression observations, and effects of CALHM5 deficiency in a mouse aneurysm model. This predates the manuscript's October 2026 submission and directly informs its vascular-smooth-muscle expression result and therapeutic discussion. Distinguish human observational evidence, cellular mechanism, and mouse intervention evidence; none alone establishes clinical efficacy. Explain what this resource profile adds beyond the existing literature.

3. Remove conclusions derived from undefined literature-similarity values and repair the underlying analysis. All ten displayed literature-neighbor scores are NaN. Nevertheless, the discussion identifies 5_8S_rRNA as the strongest neighbor and describes the scores as low. An undefined value supports neither ranking nor weak similarity. Diagnose whether this reflects identifier mismatch, inadequate source coverage, a numerical failure, or another cause. Recompute where possible; otherwise report unavailable similarity and remove the neighbor-based inference. Do not convert NaNs to zeros. Define the AEBP1 co-mention fraction and its denominator: one shared PMID and a fraction of 0.0125 cannot be interpreted from the reported two-PMID total without knowing the normalization.

4. Clarify coexpression selection and limit mechanistic interpretation. Twenty returned neighbors are described as “significantly coexpressed,” but no statistical test, sample count, null model, multiple-testing correction, or threshold is given. Specify the Geneshot endpoint, underlying ARCHS4 release, correlation/similarity definition, sample universe, and selection rule. If these are simply the twenty highest-ranked similarities, use that description rather than statistical significance. A ranked neighbor list also does not establish interactions among neighbors or a functional network. Broad expression correlations can reflect cell composition, tissue identity, or study effects. Explain those alternatives and either add a relevant within-context/independent validation analysis or keep the conclusions descriptive. The reported smooth-muscle expression makes tissue composition a particularly relevant possibility.

5. Audit tissue parsing and zero handling. The expression table mixes named cell types with hierarchical labels such as “System” and “System.Connective Tissue,” several assigned exactly zero. Establish whether these are measured tissue summaries, hierarchy nodes, or missing values represented as zero. Provide raw responses, sample counts, units/transformation, valid-label selection, and missing-value rules. Explain how the baseline mean 4.336 and SD 3.921 were obtained and why they are appropriate for this expression scale. The z-scores are numerically consistent with that formula, but do not by themselves demonstrate tissue-specific enrichment or statistical significance. Recompute the ranking using valid observations and distinguish tissues from cultured cell types.

6. Document Harmonizome score semantics and reconcile the figure caption. All ten displayed leading associations have values of −20.000, while the caption names JASPAR TF predictions, a dataset absent from the ten listed sources. Verify the actual table/image inputs and correct the caption. Supply the full association export, dataset provenance, score transformations, ranking direction, and tie handling; explain whether −20 represents a cap, transformed significance, or another quantity rather than assuming it is a conventional z-score. Harmonizome documentation describes standardized association values as dataset-dependent. Its PerturbAtlas dataset description describes expression changes in one gene following perturbation of another. Such associations do not establish physical interaction or direct regulation. Separate human and mouse evidence, document ortholog mapping, and retain study and experimental context rather than treating the aggregate count as independent biological support.

Reproducibility and editorial corrections

7. Provide an auditable analysis and repair citations/templates. Deposit code, query parameters and dates, resource versions, identifier crosswalks, and raw responses. Distinguish successful empty results from unsupported identifiers, failed requests, and invalid outputs, especially for GSFM and regulatory queries. Verify that FAM26E and CALHM5 resolve consistently and that the PDF matches this version. Audit all references: the verified Geneshot paper is Lachmann et al., 2019, and the original Harmonizome paper is Rouillard et al., 2016, unlike the supplied entries. Correct the fibrosis-specific description of CFDE using its NIH overview. Remove the unresolved symbol placeholder and duplicated family wording; reconcile the no-predictions result with the GSFM prediction caption, and escape the vertical bar inside the HRAS table entry.

A revised descriptive profile could be valuable without new wet-lab experiments if it corrects the channel biology, incorporates relevant literature, resolves invalid outputs, and makes every reported association traceable to its source.