{ "version": "M5 v6", "freshHoldout": false, "superiorityEstablished": false, "protocol": { "lockedAtUTC": "2026-10-13T15:02:50.440954+00:00", "freshHoldout": false, "superiorityEstablished": false, "changes": [ "Exclude SMG1 to isolate the UPF1 endpoint.", "Use original 103 sequence features, excluding added geometry and study-confounded perturbation flag.", "Regularize M5 to depth 1/2/3; train every family with the same precision weights." ], "weightRule": "Per training study: 1/(lfcSE squared + training median SE squared); replace invalid SE with training median; cap at training 95th percentile; normalize equal study totals. Recomputed within inner training, never from validation/test uncertainty.", "selection": "12 configurations per family; deterministic gene-disjoint 80/20 inner split; maximize mean unweighted inner Spearman; outer leave-one-study-out diagnostics. No sign flipping, test subgroup selection or outer-score selection.", "families": { "M5_v6": [ { "max_depth": 1, "min_child_weight": 20, "reg_lambda": 10 }, { "max_depth": 1, "min_child_weight": 20, "reg_lambda": 100 }, { "max_depth": 1, "min_child_weight": 100, "reg_lambda": 10 }, { "max_depth": 1, "min_child_weight": 100, "reg_lambda": 100 }, { "max_depth": 2, "min_child_weight": 20, "reg_lambda": 10 }, { "max_depth": 2, "min_child_weight": 20, "reg_lambda": 100 }, { "max_depth": 2, "min_child_weight": 100, "reg_lambda": 10 }, { "max_depth": 2, "min_child_weight": 100, "reg_lambda": 100 }, { "max_depth": 3, "min_child_weight": 20, "reg_lambda": 10 }, { "max_depth": 3, "min_child_weight": 20, "reg_lambda": 100 }, { "max_depth": 3, "min_child_weight": 100, "reg_lambda": 10 }, { "max_depth": 3, "min_child_weight": 100, "reg_lambda": 100 } ], "ExtraTrees": [ { "min_samples_leaf": 3, "max_features": 0.5, "max_depth": 12 }, { "min_samples_leaf": 3, "max_features": 0.5, "max_depth": null }, { "min_samples_leaf": 3, "max_features": 1.0, "max_depth": 12 }, { "min_samples_leaf": 3, "max_features": 1.0, "max_depth": null }, { "min_samples_leaf": 10, "max_features": 0.5, "max_depth": 12 }, { "min_samples_leaf": 10, "max_features": 0.5, "max_depth": null }, { "min_samples_leaf": 10, "max_features": 1.0, "max_depth": 12 }, { "min_samples_leaf": 10, "max_features": 1.0, "max_depth": null }, { "min_samples_leaf": 30, "max_features": 0.5, "max_depth": 12 }, { "min_samples_leaf": 30, "max_features": 0.5, "max_depth": null }, { "min_samples_leaf": 30, "max_features": 1.0, "max_depth": 12 }, { "min_samples_leaf": 30, "max_features": 1.0, "max_depth": null } ], "ridge": [ { "alpha": 0.01 }, { "alpha": 0.1 }, { "alpha": 1.0 }, { "alpha": 10.0 }, { "alpha": 100.0 }, { "alpha": 1000.0 }, { "alpha": 10000.0 }, { "alpha": 100000.0 }, { "alpha": 1000000.0 }, { "alpha": 10000000.0 }, { "alpha": 100000000.0 }, { "alpha": 1000000000.0 } ] }, "sources": { "docs/validation/2026-10-13-m5-development-v5/development-inputs.parquet": "72bc28040fd68265cede59c5f5cd81e7fe0ba58bdc4a545a88de0ad5090dd345", "docs/validation/2026-10-13-response-reconciliation-v4/Aliouat2019-te-targets.csv": "a1f733941a86ab0810ae14bd7bdf1593397d00093890b8b417fa40452930b917", "docs/validation/2026-10-13-response-reconciliation-v4/NMDRHT-te-targets.csv": "0354af35151579d37fee71e81712783f8c185f01f7ef3fa040a032455d5ed4ed" }, "features": [ "n_uorfs_aug", "n_uorfs_near", "max_uorf_len", "dist_cap_to_first_uorf", "first_uorf_is_aug", "uorf_overlaps_main_orf", "min_dist_stop_to_cds", "kozak_max", "kozak_first", "utr_len", "utr_gc", "utr_dg", "kozak_x_dist", "n_uorfs_x_utrlen", "gc_x_dg", "main_kozak", "cds_len", "utr3_len", "dg_per_nt", "utr_kmer_A", "utr_kmer_C", "utr_kmer_G", "utr_kmer_T", "utr_kmer_AA", "utr_kmer_AC", "utr_kmer_AG", "utr_kmer_AT", "utr_kmer_CA", "utr_kmer_CC", "utr_kmer_CG", "utr_kmer_CT", "utr_kmer_GA", "utr_kmer_GC", "utr_kmer_GG", "utr_kmer_GT", "utr_kmer_TA", "utr_kmer_TC", "utr_kmer_TG", "utr_kmer_TT", "utr_kmer_AAA", "utr_kmer_AAC", "utr_kmer_AAG", "utr_kmer_AAT", "utr_kmer_ACA", "utr_kmer_ACC", "utr_kmer_ACG", "utr_kmer_ACT", "utr_kmer_AGA", "utr_kmer_AGC", "utr_kmer_AGG", "utr_kmer_AGT", "utr_kmer_ATA", "utr_kmer_ATC", "utr_kmer_ATG", "utr_kmer_ATT", "utr_kmer_CAA", "utr_kmer_CAC", "utr_kmer_CAG", "utr_kmer_CAT", "utr_kmer_CCA", "utr_kmer_CCC", "utr_kmer_CCG", "utr_kmer_CCT", "utr_kmer_CGA", "utr_kmer_CGC", "utr_kmer_CGG", "utr_kmer_CGT", "utr_kmer_CTA", "utr_kmer_CTC", "utr_kmer_CTG", "utr_kmer_CTT", "utr_kmer_GAA", "utr_kmer_GAC", "utr_kmer_GAG", "utr_kmer_GAT", "utr_kmer_GCA", "utr_kmer_GCC", "utr_kmer_GCG", "utr_kmer_GCT", "utr_kmer_GGA", "utr_kmer_GGC", "utr_kmer_GGG", "utr_kmer_GGT", "utr_kmer_GTA", "utr_kmer_GTC", "utr_kmer_GTG", "utr_kmer_GTT", "utr_kmer_TAA", "utr_kmer_TAC", "utr_kmer_TAG", "utr_kmer_TAT", "utr_kmer_TCA", "utr_kmer_TCC", "utr_kmer_TCG", "utr_kmer_TCT", "utr_kmer_TGA", "utr_kmer_TGC", "utr_kmer_TGG", "utr_kmer_TGT", "utr_kmer_TTA", "utr_kmer_TTC", "utr_kmer_TTG", "utr_kmer_TTT" ], "limits": [ "Both studies already used and repeatedly inspected; development only.", "Acute 12h versus chronic 72h UPF1 depletion, unmatched NMDRHT time controls, source quantification differences persist.", "Sequence-only input cannot identify perturbation duration or experimental context.", "Precision weighting relies on approximate DESeq2 standard errors; it does not repair study confounding.", "Gene overlap across studies is reported; all eligible outer genes are retained." ] }, "audit": { "sharedGenes": 4628, "targetSpearman": -0.18859932464146273, "interpretation": "Measured endpoints disagree across studies; not proof of a sign error or specific biological cause. Coefficient construction reviewed: both target RPF response minus RNA response. No labels changed." }, "folds": [ { "heldStudy": "Aliouat2019", "genes": 5268, "overlappingGenes": 4628, "metrics": { "M5_v6": { "spearman": -0.21904729430995962, "mae": 0.41323403787325835, "r2": -0.3883784082088917 }, "ExtraTrees": { "spearman": -0.21243892464576924, "mae": 0.4318184883068632, "r2": -0.5032536110787553 }, "ridge": { "spearman": -0.17884214255739603, "mae": 0.3964359414309435, "r2": -0.31834848165593654 }, "training_median": { "spearman": null, "mae": 0.33846793721561474, "r2": -0.044244709574735586 } }, "comparisons": [ { "baseline": "ExtraTrees", "deltaSpearman": -0.006608369664190372, "adjustedPairedCi": [ -0.0313182898373483, 0.012983474694475267 ] }, { "baseline": "ridge", "deltaSpearman": -0.04020515175256359, "adjustedPairedCi": [ -0.06738125251032764, -0.014977491640156911 ] } ], "selections": { "M5_v6": { "configurationId": 8, "innerSpearman": 0.32878326245944145 }, "ExtraTrees": { "configurationId": 2, "innerSpearman": 0.35488034002688884 }, "ridge": { "configurationId": 0, "innerSpearman": 0.24349407651288743 } } }, { "heldStudy": "NMDRHT", "genes": 4886, "overlappingGenes": 4628, "metrics": { "M5_v6": { "spearman": -0.32067035788201437, "mae": 0.48198898442400856, "r2": -0.1851241954455145 }, "ExtraTrees": { "spearman": -0.3255907721144607, "mae": 0.4751745734493849, "r2": -0.15774983281660404 }, "ridge": { "spearman": -0.281211610339668, "mae": 0.4694046534453878, "r2": -0.141290464783552 }, "training_median": { "spearman": null, "mae": 0.447769565919392, "r2": -0.057090534796314873 } }, "comparisons": [ { "baseline": "ExtraTrees", "deltaSpearman": 0.004920414232446324, "adjustedPairedCi": [ -0.009745179498410977, 0.019162607259117188 ] }, { "baseline": "ridge", "deltaSpearman": -0.03945874754234635, "adjustedPairedCi": [ -0.06399073927112503, -0.018534576652131356 ] } ], "selections": { "M5_v6": { "configurationId": 11, "innerSpearman": 0.24748727182605795 }, "ExtraTrees": { "configurationId": 10, "innerSpearman": 0.23842517938761545 }, "ridge": { "configurationId": 5, "innerSpearman": 0.21223422499536068 } } } ], "finalModels": { "M5_v6": { "configurationId": 8, "sha256": "92db2649385ec4f2f9ce0b81ce6aab0d332f76052b74feab7af564b7cfb088e6" }, "ExtraTrees": { "configurationId": 2, "sha256": "d9e4f72ba77d59b7248133551fde65c0f94076d639ddeddc78fb23c847d630d8" }, "ridge": { "configurationId": 2, "sha256": "070a9f55da48f483dd7ee3b161a7ad210a533abb75930a20f3bf9b1477b0595d" } }, "previousV5": { "Aliouat2019": { "spearman": -0.3122084098213585, "mae": 0.4548342500886957, "r2": -0.5956244139223363 }, "NMDRHT": { "spearman": -0.2701988289045821, "mae": 0.5059331659600207, "r2": -0.27514061116864696 } }, "promotionGate": { "passed": false, "reason": "Negative correlation and negative R\u00b2 in both UPF1 transfer diagnostics; ridge has higher Spearman on both; no fresh study has qualified. Transfer is not fixed." }, "search": { "date": "2026-10-13", "qualifiedStudies": 0, "freshTestRun": false, "scope": "Targeted web searches using UPF1/RPF/ribosome profiling and human/HeLa/HepG2 terms, followed by primary GEO metadata and two full-text method reviews. An extension of previous audits, not an exhaustive repository census. No numerical outcomes from new studies were scored.", "decisions": [ { "id": "GSE296644", "url": "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE296644", "reason": "UPF1 knockdown with Oxford Nanopore direct RNA sequencing; no matched RPF assay established." }, { "id": "GSE246483", "url": "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE246483", "reason": "Human UPF1 knockdown/mutant complementation, triplicate rRNA-depleted RNA-seq; no matched RPF assay established." }, { "id": "GSE228561 / GSE228565", "url": "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE228561", "reason": "Human renal-cell-carcinoma ribosome profiling follows fumarate-related genotype changes. The screened mutant RPF sample does not establish UPF1 perturbation." }, { "id": "GSE252369 / NSUN methylation", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC11235314/", "reason": "Human UPF1 knockdown is measured by RNA-seq in WT and NSUN2 knockout cells; UPF1 footprints are CLIP/RNA-binding measurements, not matched ribosome-protected fragments." }, { "id": "Human heart / fibroblast RBP analysis", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC8687540/", "reason": "RNA/Ribo-seq and RBP-target correlations are available, but this is not a direct UPF1-perturbation experiment. Fibroblast replication uses TGF-beta stimulation." } ], "nextEvidence": "A new matched human UPF1 perturbation/control RNA-and-RPF study with biological replicates, exact sample pairing, perturbation duration and time-matched controls. Obtain author clarification of the existing count discrepancy and harmonize raw-read quantification before another broad transfer claim. Freeze models and analysis rules before examining any new outcomes." } }