Review-quality audit: templated-boilerplate scan

Review-quality audit: templated-boilerplate scan

Auto-generated by projects/REVIEW_QUALITY_AUDIT/scan_boilerplate.py.

Tier 1 (critical) — fake/non-specific supporting evidence

Generic placeholder supporting_text on many annotations: the evidence does not support the specific term. Same defect class as the mouse Fyn review. Re-review fully.

File reviewed placeholder uses top templated reason
none

Tier 2 — summary AND reason both templated

Both the summary and the reason are drawn from a tiny templated set (unique-summary ratio ≤ 0.15 and unique-reason ratio ≤ 0.15). The per-annotation rationale may carry no real curation signal even though the supporting_text may be a real quote. Inspect for duplicate-row artifacts before bulk re-review.

File reviewed uniq-summary ratio uniq-reason ratio top reason
human/GRB2/GRB2-ai-review.yaml 949 0.073 0.075 564× "Per curation policy, GO:0005515 protein binding is uninformative: it does not specify what"
human/SPRED1/SPRED1-ai-review.yaml 506 0.111 0.089 423× "GO:0005515 protein binding is uninformative about SPRED1 function. Removal does not disput"
human/CASP6/CASP6-ai-review.yaml 257 0.128 0.148 195× "The row comes from a neurodegeneration interactome screen and does not define a CASP6 mole"
human/SOS1/SOS1-ai-review.yaml 246 0.146 0.142 127× "SOS1 is predominantly cytosolic when inactive; consistent across the ARBA IEA and Reactome"

Tier 3 — reason-only boilerplate (low severity)

Only the one-line reason is templated; the summary is term-specific (unique-summary ratio > 0.15) and supporting_text is a real quote. The substantive review is genuine — only the reason field is lazy. Low priority; can be tightened in bulk later.

File reviewed uniq-summary ratio uniq-reason ratio top reason
human/FGFR3/FGFR3-ai-review.yaml 440 0.934 0.082 307× "Generic protein binding is uninformative about FGFR3 function; removal does not imply the"
mouse/Bcl2/Bcl2-ai-review.yaml 358 0.458 0.128 168× "Bcl2-dependent survival has broad tissue and developmental consequences; these terms are b"
DROME/N/N-ai-review.yaml 299 0.595 0.09 180× "Notch is the single Drosophila Notch receptor and its signaling is used reiteratively in c"
mouse/Edn1/Edn1-ai-review.yaml 253 0.684 0.055 114× "This overstates the direct role of the gene product; the curated model is better captured"
mouse/Pten/Pten-ai-review.yaml 231 0.662 0.035 169× "Secondary or downstream function"
mouse/Cdc42/Cdc42-ai-review.yaml 226 0.509 0.071 98× "The term directly reflects Cdc42 GDP/GTP switch activity, effector binding, membrane local"
mouse/Syk/Syk-ai-review.yaml 223 0.498 0.063 106× "The annotation is biologically plausible for Syk-dependent immune, platelet, myeloid, vasc"
rat/Hspa8/Hspa8-ai-review.yaml 208 0.635 0.087 133× "Kept as non-core to preserve potentially valid context-specific annotation without elevati"
mouse/Hsp90aa1/Hsp90aa1-ai-review.yaml 201 0.488 0.065 118× "Retain this annotation as context-specific while keeping the core review focused on ATP-de"
human/PSMA1/PSMA1-ai-review.yaml 191 0.22 0.063 86× "The reviewed core function occurs in cytosolic and nuclear proteasome pools, supported by"
mouse/Myc/Myc-ai-review.yaml 181 0.619 0.061 73× "The term directly reflects Myc DNA binding, transcription factor activity, Myc-Max complex"
human/KASH5/KASH5-ai-review.yaml 178 0.916 0.107 144× "GO:0005515 protein binding is uninformative. Removal does not imply the reported interacti"
mouse/Ednra/Ednra-ai-review.yaml 172 0.715 0.07 101× "This annotation captures a real downstream or context-specific consequence of Ednra activi"
mouse/Calm1/Calm1-ai-review.yaml 155 0.465 0.026 94× "Tissue-specific or specialized function"
human/PSMB5/PSMB5-ai-review.yaml 154 0.201 0.071 75× "The reviewed core function occurs in cytosolic and nuclear proteasome pools, supported by"
human/FRS3/FRS3-ai-review.yaml 146 1.0 0.103 101× "Generic protein binding is uninformative and the high-throughput assay does not support a"
mouse/Calm2/Calm2-ai-review.yaml 134 0.687 0.097 57× "Tissue-specific or specialized function"
human/FRS2/FRS2-ai-review.yaml 130 1.0 0.146 97× "Consistent with the experimentally established myristoyl-dependent plasma-membrane localiz"
DROME/Dl/Dl-ai-review.yaml 129 0.543 0.101 54× "Delta acts as the principal transmembrane DSL ligand for Notch in Drosophila, and its requ"
mouse/Dnmt1/Dnmt1-ai-review.yaml 127 0.346 0.087 74× "Secondary or downstream function"
mouse/Calm3/Calm3-ai-review.yaml 125 0.992 0.136 53× "Tissue-specific or specialized function"
mouse/Hras/Hras-ai-review.yaml 119 0.487 0.084 62× "This term reflects a cell-type, disease-model, developmental, transcriptional, or phenotyp"
human/AGO3/AGO3-ai-review.yaml 95 0.316 0.116 60× "The falcon report supports AGO3 function in miRISC-mediated small-RNA-guided repression an"
human/ABCA7/ABCA7-ai-review.yaml 81 0.16 0.086 55× "The term aligns with direct biochemical evidence or well-supported cellular consequences o"
human/APH1A/APH1A-ai-review.yaml 73 0.178 0.096 38× "The term aligns with APH1A function as a non-catalytic gamma-secretase subunit that suppor"
human/PSENEN/PSENEN-ai-review.yaml 71 0.155 0.099 29× "The term aligns with PSENEN/PEN-2 function as an essential non-catalytic gamma-secretase s"
human/CD33/CD33-ai-review.yaml 45 0.311 0.133 18× "This term captures the core CD33/Siglec-3 role as a myeloid cell-surface sialic-acid-bindi"
rat/Slc13a2/Slc13a2-ai-review.yaml 44 1.0 0.091 32× "This term is directly supported by Slc13a2 sodium-coupled dicarboxylate transport of succi"
human/APH1B/APH1B-ai-review.yaml 43 0.256 0.116 22× "The term aligns with APH1B function as a non-catalytic gamma-secretase subunit that suppor"
DANRE/cyp26a1/cyp26a1-ai-review.yaml 38 0.711 0.079 25× "This annotation is broad, inferred, or reflects downstream developmental/physiological con"
DANRE/gnptab/gnptab-ai-review.yaml 26 1.0 0.077 14× "This annotation is broad, inferred, or reflects downstream developmental/physiological con"