Review Papers Against Their Own Results

Visual summary — data first

What the paper did (concise, data-linked)

• Re-annotated NCBI RefSeq GRCh38.p14 and T2T-CHM13v2.0, classifying genes as coding-only, noncoding-only, or 'bifunctional' (has ≥1 coding and ≥1 noncoding transcript) and reported 4,387 bifunctional genes in GRCh38.p14 (4,333 in T2T) Insan et al., methods and counts
• Performed gene-length / exon / transcript comparisons, chromosomal/strand distributions, GO/KEGG/WikiPathway enrichments, orthology mapping (chimp, mouse, zebrafish), circRNA mapping (circAtlas), ORF prediction from nc-isoforms (ORFipy), Kozak and hydrophobicity scoring, BLASTP checks, and miRNA seed mapping (microCLIP + TargetScan) — and packaged results into a Shiny database and GitHub repository Insan et al., pipelines and outputs

Main results (data-driven bullets)

Critical appraisal — strengths

1. Comprehensive, transparent reannotation pipeline applied to two independent RefSeq assemblies (GRCh38.p14 and T2T-CHM13v2.0) with scripts and a Shiny database made public — strong reproducibility intent and data sharing Insan et al., data/code availability
2. Multi-layered evidence: assembly concordance, long-read and short-read expression visualization, RT-qPCR validation, ORF predictions with contextual scoring, circRNA mapping and miRNA-seed overlaps provide a multi-angled dataset rather than a single line of inference Insan et al., multi-modal evidence
3. Useful resource: the Shiny database and GitHub scripts are practical for the community to query candidate bifunctional genes and to design follow-up functional experiments (CRISPR, ribosome profiling, MS) per-locus Insan et al., resource availability

Critical appraisal — limitations, blindspots, and alternative interpretations

1. Annotation dependence and circularity: classification hinges on current RefSeq transcript biotypes (NM/XM vs NR/XR). Predicted transcripts (XM/XR) inflate counts relative to strictly experimentally validated transcripts; authors included predicted transcripts intentionally — this increases sensitivity but also false-positive risk for functional bifunctionality Insan et al., annotation choices
2. Expression-level and stability uncertainties: many nc-isoforms are low abundance and may be NMD substrates; detection in long-read or RT-qPCR proves transcription but not stability, subcellular localization, or function (e.g., ceRNA competition requires high stoichiometry) Insan et al., NMD and expression caveats
3. Translation predictions vs biochemical proof: ORF prediction (≥150 nt), Kozak scores and hydrophobicity are useful filters but do not demonstrate translation; while authors cross-checked OpenProt/Ribo-seq/MS where possible, only a minority of predicted ORFs will be shown to produce stable peptides — mass-spec, ribosome profiling with junction-aware mapping, or immunoprecipitation are needed to validate miP production Insan et al., ORF vs translation
4. Orthology mapping caveats: ortholog detection using gene symbols or automated pipelines may miss lineage-specific transcript evolution; showing a gene has orthologous DNA does not prove conserved bifunctionality (i.e., the ortholog may be coding-only in the other species). Authors report only 13 genes truly bifunctional across all four species — highlighting limited conservation of the bifunctional state itself Insan et al., orthology limits
5. Functional over-interpretation risk: enriched GO terms (e.g., vesicular transport, glycosylation, RNA-binding) indicate categories of genes that produce noncoding transcripts, but do not prove the noncoding isoform contributes functionally to those processes; enrichment may reflect that large, highly-spliced gene families (e.g., nucleoporins, SRSFs) are annotation-rich and so biased toward bifunctional assignment Insan et al., interpretation caution

What would strengthen / falsify the key claims (actionable next steps)

1. Ribosome profiling (Ribo-seq) with junction-aware mapping across tissues/cell lines to detect translation of predicted smORFs from nc-isoforms, paired with targeted mass-spectrometry (PRM/SRM) for candidate microproteins — necessary to move predictions to validated microproteome entries Insan et al., need for Ribo-seq/MS
2. CRISPR-based isoform-specific perturbations (splice-site editing, exon deletion, CRISPRi targeted to nc-exon promoters) combined with transcriptome and proteome readouts to test functional effects of nc-isoforms on cognate mRNA, protein levels, and cell phenotype (rescue with isoform-only constructs) — will directly test ceRNA or protein-antagonist hypotheses.
3. Quantitative ceRNA tests: measure absolute copies per cell (smFISH or Nanostring) for mRNA and nc-isoform plus miRNA abundance; perform titration experiments to see if nc-isoform levels alter miRNA-mediated repression of coding isoforms — because computational seed-site overlaps are insufficient to claim ceRNA functions Insan et al., ceRNA caution
4. Cross-species functional tests: for the 13 conserved bifunctional genes (many splicing/RNA-binding factors) test whether the noncoding isoforms in other vertebrates are expressed and functional (e.g., reporter assays, cross-complementation) — would validate evolutionary conservation of bifunctionality as functionally meaningful Insan et al., cross-species list

Bottom-line assessment (concise)

The paper is a technically careful, well-documented genome-wide reannotation and resource that compellingly catalogs candidate bifunctional loci and generates many experimentally testable predictions. Its primary limitation is that many functional claims (translation of predicted ORFs, ceRNA activity, regulatory impact of nc-isoforms) remain predictions and need locus-level biochemical validation; however the database and code substantially lower the barrier to those critical follow-ups Insan et al., conclusion and caveats