{
  "date": "2026-10-13",
  "endpoint": "SMG1 inhibition versus DMSO in WT K562 cells: log2 RPF change minus log2 RNA change; separate pathway transfer, not UPF1-depletion validation",
  "protocol": {
    "date": "2026-10-13",
    "lockedAtUTC": "2026-10-13T14:17:25.372416+00:00",
    "endpoint": "SMG1 inhibition versus DMSO in WT K562 cells: log2 RPF change minus log2 RNA change; separate pathway transfer, not UPF1-depletion validation",
    "study": "Zinshteyn 2021 / PRJNA675773 / PMC8555832",
    "samples": [
      "WT_DMSO_1_RNA",
      "WT_DMSO_2_RNA",
      "WT_SMG1i_1_RNA",
      "WT_SMG1i_2_RNA",
      "WT_DMSO_1_mono",
      "WT_DMSO_2_mono",
      "WT_SMG1i_1_mono",
      "WT_SMG1i_2_mono"
    ],
    "replication": "Two biological lysates per condition; RNA and monosome assays paired by condition/replicate. Knockout and disome libraries excluded.",
    "source": "https://journals.plos.org/plosgenetics/article/file?id=10.1371/journal.pgen.1009813.s013&type=supplementary",
    "modelSha256": {
      "reconciled_v4": "8de9d75816d11b17cd03a0988fb4cc1b44cf5958dc01750afa60c05d76db3aee",
      "ExtraTrees": "48cc22c7316e546c4a294fc016bceac8b643be47a50139b4066292e59cca00fc",
      "ridge": "65d3ab7c933060c0198ab0388a85915a2dd904f27d56421d35721065b403370e"
    },
    "featureBankSha256": "55d4c68114559046b464fc6834d00d3c985f6728c97957ccc325c9affc071e56",
    "predictionSha256": "c1758a16010db655eb4826a9c102a6c82073a14b5e2b7b4a1dec32b2141316da",
    "predictBeforeOutcomes": true,
    "featureBankRows": 6159,
    "mapping": "Use supplied gene_id, strip Ensembl version, retain ENSG identifiers with unique tx_id-to-gene mapping; sum transcript counts per gene before rounding. Exclude ambiguous transcript mappings and duplicate tx_id rows, no isoform selection based on outcomes.",
    "countFilter": "At least 10 counts in >=2 of 4 samples per assay, applied across full mapped count matrix before intersection with feature bank.",
    "estimator": "DESeq2 paired design ~sample + assay + treatedRPF, separately estimated median-ratio size factors within assay. Coefficient treatedRPF. Exclude nonconverged/nonfinite targets; no p-value or response-magnitude selection; no fit/predictor tuning.",
    "primaryMetric": "Spearman rank correlation",
    "uncertainty": "2000 paired whole-gene bootstrap resamples, seed 20261013. Two predeclared primary candidate-minus-baseline Spearman comparisons, Bonferroni-adjusted 97.5% intervals. Marginal 95% metric intervals descriptive.",
    "secondaryMetrics": [
      "MAE",
      "R2",
      "training-median baseline"
    ],
    "superiorityRule": "Both adjusted Spearman difference intervals entirely >0 supports ranking advantage only on this SMG1 cohort. No general, UPF1-specific or therapeutic superiority. Report absolute performance and all comparisons.",
    "runtime": "30 rotating-order warm rounds for batches 1/32 sorted by gene, same four-thread budget. Precomputed-feature input validation/copy, imputation, prediction and JSON serialization; no sequence folding, feature extraction, HTTP, or target estimation. Load and first all-bank inference recorded separately.",
    "independence": "New K562 SMG1 study distinct from Aliouat HCT116 and Boehm NMDRHT; publication/accession/sample metadata checked. Raw read identity not deduplicated. Gene overlap expected and reported. No independent third-party custody. Supplement header/methods inspected previously, not numeric gene responses.",
    "limitations": [
      "Different perturbation and cell context from training; SMG1 blockade is not UPF1 depletion.",
      "GENCODE35 source quantification versus existing representative transcript features; gene aggregation and annotation versions differ.",
      "Feature bank was originally selected from NMDRHT response availability and limits coverage.",
      "Two replicates per condition; gene bootstrap does not quantify cross-study or biological-replicate uncertainty.",
      "No retraining, calibration, sign flipping, winning subgroup selection or serving-model promotion."
    ],
    "trainingMedian": -0.042195069063600155
  },
  "mapping": {
    "sourceRows": 24244,
    "invalidGeneRows": 1,
    "duplicateTranscriptRowsExcluded": 0,
    "mappedTranscripts": 24243,
    "mappedGenes": 12922,
    "sourceSha256": "b71d9016b59d53066b4c0da3269ceb3d5950c880e11631fcb76da9dd81a320b1",
    "sourceRetrievedAtUTC": "2026-10-13T14:18:03.663892+00:00",
    "columns": [
      "WT_DMSO_1_RNA",
      "WT_DMSO_2_RNA",
      "WT_SMG1i_1_RNA",
      "WT_SMG1i_2_RNA",
      "WT_DMSO_1_mono",
      "WT_DMSO_2_mono",
      "WT_SMG1i_1_mono",
      "WT_SMG1i_2_mono"
    ]
  },
  "estimator": {
    "inputGenes": 12922,
    "countFilterPassed": 9539,
    "finiteTargets": 9539,
    "nonConverged": 0,
    "design": "~ sample + assay + treatedRPF",
    "coefficient": "treatedRPF",
    "sampleNames": [
      "WT_DMSO_1_RNA",
      "WT_DMSO_2_RNA",
      "WT_SMG1i_1_RNA",
      "WT_SMG1i_2_RNA",
      "WT_DMSO_1_mono",
      "WT_DMSO_2_mono",
      "WT_SMG1i_1_mono",
      "WT_SMG1i_2_mono"
    ],
    "pairs": [
      "WT_DMSO_1",
      "WT_DMSO_2",
      "WT_SMG1i_1",
      "WT_SMG1i_2",
      "WT_DMSO_1",
      "WT_DMSO_2",
      "WT_SMG1i_1",
      "WT_SMG1i_2"
    ],
    "assay": [
      "RNA",
      "RNA",
      "RNA",
      "RNA",
      "RPF",
      "RPF",
      "RPF",
      "RPF"
    ],
    "sizeFactors": [
      1.0133,
      1.041,
      1.0668,
      0.8932,
      1.2062,
      1.1518,
      0.9946,
      0.7325
    ],
    "designRank": 6,
    "residualDf": 2,
    "packageVersion": "1.50.2",
    "filter": "At least 10 counts in at least two samples per assay; no significance filter",
    "normalization": "Separate assay-specific median-ratio size factors",
    "target": "SMG1i minus DMSO RPF change minus RNA change",
    "limitations": "Two biological replicates per condition. Salmon estimated counts rounded after gene aggregation. No effective-length offsets or raw-read re-quantification."
  },
  "nEvaluated": 3462,
  "finiteTargetGenes": 9539,
  "featureBankGenesWithoutEligibleTarget": 2697,
  "targetGenesWithoutFeatures": 6077,
  "seenDevelopmentGenes": 3397,
  "unseenDevelopmentGenes": 65,
  "metrics": {
    "reconciled_v4": {
      "spearman": 0.19432578024320749,
      "mae": 0.26654189620826674,
      "r2": 0.016708586651904667
    },
    "ExtraTrees": {
      "spearman": 0.2558617516623866,
      "mae": 0.2619689789563071,
      "r2": 0.04472478948848313
    },
    "ridge": {
      "spearman": 0.11010393483233501,
      "mae": 0.2706385144145103,
      "r2": 0.003004540735950023
    },
    "training_median": {
      "spearman": null,
      "mae": 0.2706529413686233,
      "r2": -0.0007745234327523054
    }
  },
  "metricSpearmanCi95": {
    "reconciled_v4": [
      0.16196279606029557,
      0.22631208909911982
    ],
    "ExtraTrees": [
      0.22364986363591605,
      0.28696708652408254
    ],
    "ridge": [
      0.07857251282459418,
      0.14183481822159744
    ]
  },
  "comparisons": [
    {
      "baseline": "ExtraTrees",
      "deltaSpearman": -0.06153597141917913,
      "adjustedPairedCi": [
        -0.09624515150512632,
        -0.025666063636742013
      ],
      "supportsRankingAdvantage": false
    },
    {
      "baseline": "ridge",
      "deltaSpearman": 0.08422184541087248,
      "adjustedPairedCi": [
        0.05129635961699771,
        0.11633521851428537
      ],
      "supportsRankingAdvantage": true
    }
  ],
  "rankingAdvantageOverBoth": false,
  "runtime": [
    {
      "model": "reconciled_v4",
      "batchSize": 1,
      "warmMedianMs": 0.8590420000000876,
      "warmP95Ms": 1.128412200000328,
      "rounds": 30
    },
    {
      "model": "ExtraTrees",
      "batchSize": 1,
      "warmMedianMs": 26.023437500000135,
      "warmP95Ms": 27.132691700000098,
      "rounds": 30
    },
    {
      "model": "ridge",
      "batchSize": 1,
      "warmMedianMs": 1.0135209999999617,
      "warmP95Ms": 1.2312228000000314,
      "rounds": 30
    },
    {
      "model": "reconciled_v4",
      "batchSize": 32,
      "warmMedianMs": 0.888895500000153,
      "warmP95Ms": 1.1922857500000104,
      "rounds": 30
    },
    {
      "model": "ExtraTrees",
      "batchSize": 32,
      "warmMedianMs": 26.469417000000384,
      "warmP95Ms": 26.863129650000104,
      "rounds": 30
    },
    {
      "model": "ridge",
      "batchSize": 32,
      "warmMedianMs": 1.0489584999997525,
      "warmP95Ms": 1.1794501499997612,
      "rounds": 30
    }
  ],
  "machine": {
    "platform": "macOS-27.0.1-arm64-arm-64bit",
    "python": "3.9.6",
    "cpu": "arm"
  },
  "directUpf1Holdout": "No fresh direct human UPF1 RNA/RPF study qualified in the continued targeted search.",
  "noPromotion": true,
  "headline": "M5 beats ridge on SMG1 transfer; ExtraTrees is stronger.",
  "interpretation": "A supported ranking advantage over ridge on this cohort, not superiority over both baselines. This tests transfer from UPF1-trained models to SMG1 inhibition, not an independent direct UPF1 response.",
  "continuedSearch": {
    "recordsReturned": 52,
    "scope": "Targeted Europe PMC search of 2025\u20132026 literature using UPF1/RENT1 and ribosome-profiling/Ribo-seq terms, plus targeted web queries. Metadata screening is not full qualification of every record.",
    "qualifiedDirectUpf1Studies": 0,
    "decisions": [
      {
        "title": "TDP-43 loss / cryptic polyadenylation, 2025",
        "finding": "Methods explicitly restrict the referenced dual TDP-43/UPF1 experiment\u2019s RNA-seq to TDP-43/control and control/control arms. A matched UPF1 RNA/RPF experiment is not established.",
        "source": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12586162/"
      },
      {
        "title": "ReLiC human RNA-linked CRISPR screen, 2025",
        "finding": "Abstract describes barcoded reporters and polysome fractionation, not an established genome-wide matched UPF1 RNA/RPF contrast. Full-text API returned an access error; remains unqualified.",
        "source": "https://pubmed.ncbi.nlm.nih.gov/40442371/"
      }
    ],
    "limitation": "No new direct human UPF1 holdout qualified in this extension. This does not prove none exists; supplementary and unindexed sources may still be missed."
  },
  "additionalLimitations": [
    "Most evaluated genes were seen during model development: study independence does not mean gene independence. The 65 unseen genes are not used to select or claim a winning subgroup.",
    "The original study\u2019s selected transcript annotation used pooled ribosome data. Aggregation uses all supplied transcript rows for each gene, with no outcome-based selection here.",
    "The paired count-model coefficient is an estimate with two biological replicates per condition and two residual degrees of freedom; gene-bootstrap intervals condition on those estimates.",
    "Runtime covers precomputed-feature validation, preprocessing, inference and JSON serialization only. It is not sequence-to-result, HTTP, cold-process or production latency; thread configuration is fixed rather than optimized per method.",
    "A constant training-median baseline has undefined Spearman correlation. An initial export failed on NaN; this was fixed without altering the cohort, targets or frozen predictions."
  ]
}