{ "date": "2026-10-13", "models": [ { "id": "m4-measured-decay-v1", "key": "m4", "title": "M4 \u00b7 measured degradation-rate response", "report": { "modelId": "m4-measured-decay-v1", "endpoint": "Source-specific log2 fold change in degradation rate", "source": "https://nmdrht.uni-koeln.de/", "targetColumn": "L2FC_kdeg_tx", "trainingRows": 11669, "trainingGenes": 7597, "validationRows": 2874, "validationGenes": 1886, "selectedTestRows": 1154, "evaluatedRows": 576, "evaluatedGenes": 356, "missingTestTargets": 578, "metrics": { "response model": { "spearman": 0.5299886630917634, "mae": 0.3871924081071679, "r2": 0.30858238903452906 }, "tuned ExtraTrees": { "spearman": 0.5319684231261422, "mae": 0.381629829769887, "r2": 0.33451333537441863 }, "tuned ridge": { "spearman": 0.48987051967950174, "mae": 0.39902318092899997, "r2": 0.26981921291526034 }, "training median": { "spearman": null, "mae": 0.4518861100916346, "r2": -0.003181166922445078 }, "fitted 50-nt rule": { "spearman": 0.23239616082677841, "mae": 0.4520786616725265, "r2": 0.06945699150443985 } }, "metricCi95": { "response model": [ 0.4485771327760984, 0.5986628921943269 ], "tuned ExtraTrees": [ 0.45144907896200376, 0.5960715904967171 ], "tuned ridge": [ 0.4043620535411702, 0.5609370553625896 ], "fitted 50-nt rule": [ 0.1465231591981341, 0.3126285202684462 ] }, "comparisons": [ { "baseline": "tuned ExtraTrees", "deltaSpearman": -0.00197976003437883, "adjustedPairedCi": [ -0.02265755484767067, 0.022888920764103732 ], "supportsSuperiority": false }, { "baseline": "tuned ridge", "deltaSpearman": 0.040118143412261664, "adjustedPairedCi": [ -0.0036849024095507166, 0.08996075956133114 ], "supportsSuperiority": false }, { "baseline": "fitted 50-nt rule", "deltaSpearman": 0.297592502264985, "adjustedPairedCi": [ 0.18075998574841332, 0.4012012164431857 ], "supportsSuperiority": true } ], "allStrongBaselinesBeaten": false, "testChromosome": "chr8", "testGeneOverlap": 0, "independentExperimentalStudy": false, "limitations": [ "Fresh gene/chromosome holdout for new measured-response endpoints within the same published study, not an independent experimental study", "Published workbook was previously inspected for schema, counts and five example rows; not represented as a newly collected or never-downloaded dataset", "Previous annotation/basal-TE models may have seen gene inputs; their weights are not used to fit these response models", "No therapeutic rescue or patentability inference; negative baseline comparisons remain visible" ] }, "mapping": { "sourceRows": 45757, "nativeGtfTranscripts": 45757, "acceptedMappedInputs": 34455, "exclusions": { "noncoding_annotation": 10937, "Annotated CDS endpoint / stop-distance disagreement": 353, "nonprimary_chromosome": 12 }, "matching": "Exact native transcript/gene/coordinates and exon/CDS/UTR lengths, with independently reconstructed source stop-end distance", "features": [ "nexon", "tx_len", "cds_len", "utr5_len", "utr3_len", "stop_to_lastEJ", "downstream_junctions", "stop_upstream_50nt", "max_exon_len", "mean_exon_len", "last_exon_len", "cds_fraction", "GC_tx", "GC_utr5", "GC_cds", "GC_utr3", "utr3_mfe_nt" ], "responseColumnsUsedAsPredictors": [] }, "exampleReference": "NMDRHT00364392", "rows": [ { "method": "response model", "spearman": 0.5299886630917634, "mae": 0.3871924081071679, "r2": 0.30858238903452906, "ci95": [ 0.4485771327760984, 0.5986628921943269 ], "delta": null, "adjustedCi": null }, { "method": "tuned ExtraTrees", "spearman": 0.5319684231261422, "mae": 0.381629829769887, "r2": 0.33451333537441863, "ci95": [ 0.45144907896200376, 0.5960715904967171 ], "delta": -0.00197976003437883, "adjustedCi": [ -0.02265755484767067, 0.022888920764103732 ] }, { "method": "tuned ridge", "spearman": 0.48987051967950174, "mae": 0.39902318092899997, "r2": 0.26981921291526034, "ci95": [ 0.4043620535411702, 0.5609370553625896 ], "delta": 0.040118143412261664, "adjustedCi": [ -0.0036849024095507166, 0.08996075956133114 ] }, { "method": "training median", "spearman": null, "mae": 0.4518861100916346, "r2": -0.003181166922445078, "ci95": null, "delta": null, "adjustedCi": null }, { "method": "fitted 50-nt rule", "spearman": 0.23239616082677841, "mae": 0.4520786616725265, "r2": 0.06945699150443985, "ci95": [ 0.1465231591981341, 0.3126285202684462 ], "delta": 0.297592502264985, "adjustedCi": [ 0.18075998574841332, 0.4012012164431857 ] } ], "metadata": { "modelId": "m4-measured-decay-v1", "target": "L2FC_kdeg_tx", "predictionUnit": "Source-specific log2 fold change in degradation rate", "features": [ "nexon", "tx_len", "cds_len", "utr5_len", "utr3_len", "stop_to_lastEJ", "downstream_junctions", "stop_upstream_50nt", "max_exon_len", "mean_exon_len", "last_exon_len", "cds_fraction", "GC_tx", "GC_utr5", "GC_cds", "GC_utr3", "utr3_mfe_nt" ], "trainingMedians": { "nexon": 9.0, "tx_len": 2803.0, "cds_len": 1143.0, "utr5_len": 187.0, "utr3_len": 1062.0, "stop_to_lastEJ": -114.0, "downstream_junctions": 0.0, "stop_upstream_50nt": 0.0, "max_exon_len": 1195.0, "mean_exon_len": 264.0, "last_exon_len": 1114.0, "cds_fraction": 0.46804123711340206, "GC_tx": 0.51303441084463, "GC_utr5": 0.687974732226987, "GC_cds": 0.537926235212248, "GC_utr3": 0.446540880503145, "utr3_mfe_nt": -0.285660058027079 }, "trainingRows": 11669, "trainingGenes": 7597, "validationRows": 2874, "validationGenes": 1886, "selectedConfiguration": "leaves=15, L2=1", "validationSpearman": 0.5413544404967916, "modelSha256": "8536a26631bf8112bf6f5e8c35a0b0434a09c716de1c35d647ce999950d26ce8", "protocolSha256": "7f8d4c0d616685ad76bb6b55a9087ba79ca34a4c06501d529f2362fec54f980e", "experimentalResponseModel": true, "scope": "Published N-AID-UPF1 12h response context; not therapeutic efficacy, not biological NMD probability" }, "tuning": [ { "family": "response model", "configuration": "leaves=15, L2=1", "validationSpearman": 0.5413544404967916, "seconds": 0.4947551249999975 }, { "family": "response model", "configuration": "leaves=15, L2=10", "validationSpearman": 0.5407474628231372, "seconds": 0.4971823329999978 }, { "family": "response model", "configuration": "leaves=31, L2=1", "validationSpearman": 0.530965340414987, "seconds": 1.0191047080000004 }, { "family": "response model", "configuration": "leaves=31, L2=10", "validationSpearman": 0.5331654157781369, "seconds": 1.030771124999994 }, { "family": "response model", "configuration": "leaves=63, L2=1", "validationSpearman": 0.5189885731249306, "seconds": 1.6988347500000032 }, { "family": "response model", "configuration": "leaves=63, L2=10", "validationSpearman": 0.5210250602856688, "seconds": 2.3475784999999973 }, { "family": "tuned ExtraTrees", "configuration": "leaf=2, fraction=0.5", "validationSpearman": 0.5278215739804833, "seconds": 1.5597300000000018 }, { "family": "tuned ExtraTrees", "configuration": "leaf=2, fraction=1", "validationSpearman": 0.5254318234100679, "seconds": 0.3346594579999973 }, { "family": "tuned ExtraTrees", "configuration": "leaf=5, fraction=0.5", "validationSpearman": 0.5377223656690863, "seconds": 1.0972961659999996 }, { "family": "tuned ExtraTrees", "configuration": "leaf=5, fraction=1", "validationSpearman": 0.5119726618537273, "seconds": 0.2769534169999943 }, { "family": "tuned ExtraTrees", "configuration": "leaf=10, fraction=0.5", "validationSpearman": 0.5383730391785887, "seconds": 0.8283190000000005 }, { "family": "tuned ExtraTrees", "configuration": "leaf=10, fraction=1", "validationSpearman": 0.5050627068129558, "seconds": 0.2756331670000023 }, { "family": "tuned ridge", "configuration": "alpha=0.1", "validationSpearman": 0.512974357226172, "seconds": 0.04522495900000223 }, { "family": "tuned ridge", "configuration": "alpha=1", "validationSpearman": 0.5129829936906355, "seconds": 0.02459041700000597 }, { "family": "tuned ridge", "configuration": "alpha=10", "validationSpearman": 0.5129251715879655, "seconds": 0.03938095800000241 }, { "family": "tuned ridge", "configuration": "alpha=100", "validationSpearman": 0.5122338904710155, "seconds": 0.035838708 }, { "family": "tuned ridge", "configuration": "alpha=1000", "validationSpearman": 0.5080524520516831, "seconds": 0.048932374999999695 }, { "family": "tuned ridge", "configuration": "alpha=10000", "validationSpearman": 0.49978277269875077, "seconds": 0.05101754200000386 } ], "conclusion": "Measured-response Spearman 0.530. Higher than the fitted 50-nt rule; no established superiority over tuned ExtraTrees or ridge." }, { "id": "m5-measured-te-response-v1", "key": "m5", "title": "M5 \u00b7 measured translation-efficiency response", "report": { "modelId": "m5-measured-te-response-v1", "endpoint": "Source-specific log2 fold change in translation efficiency", "source": "https://nmdrht.uni-koeln.de/", "targetColumn": "L2FC_TE", "trainingRows": 4397, "trainingGenes": 4397, "validationRows": 1102, "validationGenes": 1102, "selectedTestRows": 660, "evaluatedRows": 186, "evaluatedGenes": 186, "missingTestTargets": 474, "metrics": { "response model": { "spearman": 0.26519868151746706, "mae": 0.5606544848324178, "r2": -0.00995011316848804 }, "tuned ExtraTrees": { "spearman": 0.3320944197604539, "mae": 0.5562700174922102, "r2": 0.019758943771714277 }, "tuned ridge": { "spearman": 0.2109861111758645, "mae": 0.5742845911143115, "r2": -0.017785656703779518 }, "training median": { "spearman": null, "mae": 0.5450986997914697, "r2": -0.011774206912628093 } }, "metricCi95": { "response model": [ 0.1150796725308795, 0.3942041622365268 ], "tuned ExtraTrees": [ 0.19489491709434448, 0.4641643837094768 ], "tuned ridge": [ 0.08407610456004472, 0.34290411838090906 ] }, "comparisons": [ { "baseline": "tuned ExtraTrees", "deltaSpearman": -0.06689573824298684, "adjustedPairedCi": [ -0.17120898070245813, 0.035342990488183304 ], "supportsSuperiority": false }, { "baseline": "tuned ridge", "deltaSpearman": 0.05421257034160257, "adjustedPairedCi": [ -0.10138674273756378, 0.2120744452241922 ], "supportsSuperiority": false } ], "allStrongBaselinesBeaten": false, "testChromosome": "chr8", "testGeneOverlap": 0, "independentExperimentalStudy": false, "limitations": [ "Fresh gene/chromosome holdout for new measured-response endpoints within the same published study, not an independent experimental study", "Published workbook was previously inspected for schema, counts and five example rows; not represented as a newly collected or never-downloaded dataset", "Previous annotation/basal-TE models may have seen gene inputs; their weights are not used to fit these response models", "No therapeutic rescue or patentability inference; negative baseline comparisons remain visible" ] }, "mapping": { "geneTableRows": 62649, "mappedEligibleGenes": 19359, "candidateTranscriptExclusions": { "missing_source_CDS_or_UTR": 16772, "invalid_or_unsupported_sequence_geometry": 13077 }, "matching": "Exact GENCODE.v42 gene/transcript FASTA and header UTR/CDS positions; representative longest CDS then transcript ID, before outcome access", "responseColumnsUsedAsPredictors": [] }, "exampleReference": "ENST00000308811.8", "rows": [ { "method": "response model", "spearman": 0.26519868151746706, "mae": 0.5606544848324178, "r2": -0.00995011316848804, "ci95": [ 0.1150796725308795, 0.3942041622365268 ], "delta": null, "adjustedCi": null }, { "method": "tuned ExtraTrees", "spearman": 0.3320944197604539, "mae": 0.5562700174922102, "r2": 0.019758943771714277, "ci95": [ 0.19489491709434448, 0.4641643837094768 ], "delta": -0.06689573824298684, "adjustedCi": [ -0.17120898070245813, 0.035342990488183304 ] }, { "method": "tuned ridge", "spearman": 0.2109861111758645, "mae": 0.5742845911143115, "r2": -0.017785656703779518, "ci95": [ 0.08407610456004472, 0.34290411838090906 ], "delta": 0.05421257034160257, "adjustedCi": [ -0.10138674273756378, 0.2120744452241922 ] }, { "method": "training median", "spearman": null, "mae": 0.5450986997914697, "r2": -0.011774206912628093, "ci95": null, "delta": null, "adjustedCi": null } ], "metadata": { "modelId": "m5-measured-te-response-v1", "target": "L2FC_TE", "predictionUnit": "Source-specific log2 fold change in translation efficiency", "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" ], "trainingMedians": { "n_uorfs_aug": 0.0, "n_uorfs_near": 10.0, "max_uorf_len": 0.0, "dist_cap_to_first_uorf": 5.0, "first_uorf_is_aug": 0.0, "uorf_overlaps_main_orf": 1.0, "min_dist_stop_to_cds": 12.0, "kozak_max": 0.0, "kozak_first": 0.5, "utr_len": 143.0, "utr_gc": 0.6645, "utr_dg": -53.5, "kozak_x_dist": 0.0, "n_uorfs_x_utrlen": 0.0, "gc_x_dg": -34.67193, "main_kozak": 0.5, "cds_len": 1452.0, "utr3_len": 939.0, "dg_per_nt": -0.36774192317839594, "utr_kmer_A": 0.16666666666666666, "utr_kmer_C": 0.3132530120481928, "utr_kmer_G": 0.3333333333333333, "utr_kmer_T": 0.16632016632016633, "utr_kmer_AA": 0.0273972602739726, "utr_kmer_AC": 0.037914691943127965, "utr_kmer_AG": 0.07526881720430108, "utr_kmer_AT": 0.013986013986013986, "utr_kmer_CA": 0.05150214592274678, "utr_kmer_CC": 0.09821428571428571, "utr_kmer_CG": 0.08333333333333333, "utr_kmer_CT": 0.06763285024154589, "utr_kmer_GA": 0.061788617886178863, "utr_kmer_GC": 0.11475409836065574, "utr_kmer_GG": 0.1056701030927835, "utr_kmer_GT": 0.041666666666666664, "utr_kmer_TA": 0.0125, "utr_kmer_TC": 0.05357142857142857, "utr_kmer_TG": 0.05785123966942149, "utr_kmer_TT": 0.029411764705882353, "utr_kmer_AAA": 0.0, "utr_kmer_AAC": 0.00501002004008016, "utr_kmer_AAG": 0.010452961672473868, "utr_kmer_AAT": 0.0, "utr_kmer_ACA": 0.005128205128205128, "utr_kmer_ACC": 0.011560693641618497, "utr_kmer_ACG": 0.005037783375314861, "utr_kmer_ACT": 0.006896551724137931, "utr_kmer_AGA": 0.01279317697228145, "utr_kmer_AGC": 0.023255813953488372, "utr_kmer_AGG": 0.020942408376963352, "utr_kmer_AGT": 0.0087527352297593, "utr_kmer_ATA": 0.0, "utr_kmer_ATC": 0.00340522133938706, "utr_kmer_ATG": 0.0, "utr_kmer_ATT": 0.0, "utr_kmer_CAA": 0.004987531172069825, "utr_kmer_CAC": 0.010309278350515464, "utr_kmer_CAG": 0.024416135881104035, "utr_kmer_CAT": 0.0019646365422396855, "utr_kmer_CCA": 0.015384615384615385, "utr_kmer_CCC": 0.02564102564102564, "utr_kmer_CCG": 0.027131782945736434, "utr_kmer_CCT": 0.019138755980861243, "utr_kmer_CGA": 0.007874015748031496, "utr_kmer_CGC": 0.02586206896551724, "utr_kmer_CGG": 0.03, "utr_kmer_CGT": 0.005780346820809248, "utr_kmer_CTA": 0.0018450184501845018, "utr_kmer_CTC": 0.02, "utr_kmer_CTG": 0.02586206896551724, "utr_kmer_CTT": 0.00980392156862745, "utr_kmer_GAA": 0.010101010101010102, "utr_kmer_GAC": 0.012244897959183673, "utr_kmer_GAG": 0.02564102564102564, "utr_kmer_GAT": 0.0028735632183908046, "utr_kmer_GCA": 0.015810276679841896, "utr_kmer_GCC": 0.034482758620689655, "utr_kmer_GCG": 0.02997275204359673, "utr_kmer_GCT": 0.021660649819494584, "utr_kmer_GGA": 0.022727272727272728, "utr_kmer_GGC": 0.03488372093023256, "utr_kmer_GGG": 0.02877697841726619, "utr_kmer_GGT": 0.011029411764705883, "utr_kmer_GTA": 0.0, "utr_kmer_GTC": 0.011363636363636364, "utr_kmer_GTG": 0.013574660633484163, "utr_kmer_GTT": 0.006060606060606061, "utr_kmer_TAA": 0.0, "utr_kmer_TAC": 0.0, "utr_kmer_TAG": 0.003048780487804878, "utr_kmer_TAT": 0.0, "utr_kmer_TCA": 0.005917159763313609, "utr_kmer_TCC": 0.019230769230769232, "utr_kmer_TCG": 0.008264462809917356, "utr_kmer_TCT": 0.010526315789473684, "utr_kmer_TGA": 0.008888888888888889, "utr_kmer_TGC": 0.016666666666666666, "utr_kmer_TGG": 0.01694915254237288, "utr_kmer_TGT": 0.006644518272425249, "utr_kmer_TTA": 0.0, "utr_kmer_TTC": 0.00966183574879227, "utr_kmer_TTG": 0.0075, "utr_kmer_TTT": 0.003484320557491289 }, "trainingRows": 4397, "trainingGenes": 4397, "validationRows": 1102, "validationGenes": 1102, "selectedConfiguration": "leaves=31, L2=10", "validationSpearman": 0.23098982733558485, "modelSha256": "75c3f257cd628aa8d17fba5a7a8c204588dc21d89acd9888d138c1917f82ee5f", "protocolSha256": "7f8d4c0d616685ad76bb6b55a9087ba79ca34a4c06501d529f2362fec54f980e", "experimentalResponseModel": true, "scope": "Published N-AID-UPF1 12h response context; not therapeutic efficacy, not biological NMD probability" }, "tuning": [ { "family": "response model", "configuration": "leaves=15, L2=1", "validationSpearman": 0.22303884339431682, "seconds": 0.9281134169999632 }, { "family": "response model", "configuration": "leaves=15, L2=10", "validationSpearman": 0.22641437217495375, "seconds": 0.8528275840000106 }, { "family": "response model", "configuration": "leaves=31, L2=1", "validationSpearman": 0.19344438654864496, "seconds": 1.5233585830000038 }, { "family": "response model", "configuration": "leaves=31, L2=10", "validationSpearman": 0.23098982733558485, "seconds": 1.570711082999992 }, { "family": "response model", "configuration": "leaves=63, L2=1", "validationSpearman": 0.20418917900734274, "seconds": 2.8519745419999936 }, { "family": "response model", "configuration": "leaves=63, L2=10", "validationSpearman": 0.21469103432398162, "seconds": 2.9828762079999933 }, { "family": "tuned ExtraTrees", "configuration": "leaf=2, fraction=0.5", "validationSpearman": 0.24523451790132697, "seconds": 2.4435432920000153 }, { "family": "tuned ExtraTrees", "configuration": "leaf=2, fraction=1", "validationSpearman": 0.15695603123545052, "seconds": 0.22522741699998505 }, { "family": "tuned ExtraTrees", "configuration": "leaf=5, fraction=0.5", "validationSpearman": 0.2579761688419504, "seconds": 1.5173062920000007 }, { "family": "tuned ExtraTrees", "configuration": "leaf=5, fraction=1", "validationSpearman": 0.12527894859659944, "seconds": 0.23384170799999993 }, { "family": "tuned ExtraTrees", "configuration": "leaf=10, fraction=0.5", "validationSpearman": 0.24807773341143913, "seconds": 1.1602293340000074 }, { "family": "tuned ExtraTrees", "configuration": "leaf=10, fraction=1", "validationSpearman": 0.1250121777364734, "seconds": 0.2134272079999846 }, { "family": "tuned ridge", "configuration": "alpha=0.1", "validationSpearman": 0.2084205691424611, "seconds": 0.03975333299996464 }, { "family": "tuned ridge", "configuration": "alpha=1", "validationSpearman": 0.2069423585520059, "seconds": 0.051110542000003534 }, { "family": "tuned ridge", "configuration": "alpha=10", "validationSpearman": 0.2053266243599043, "seconds": 0.1221907919999694 }, { "family": "tuned ridge", "configuration": "alpha=100", "validationSpearman": 0.20141163578881321, "seconds": 0.06447579100000667 }, { "family": "tuned ridge", "configuration": "alpha=1000", "validationSpearman": 0.19653578433024596, "seconds": 0.08038095800003475 }, { "family": "tuned ridge", "configuration": "alpha=10000", "validationSpearman": 0.17986498629150985, "seconds": 0.06556079099999579 } ], "conclusion": "Measured TE-response Spearman 0.265. No established superiority over tuned ExtraTrees or ridge. Absolute MAE is worse than the training-median predictor and R\u00b2 is negative; predictions remain exploratory." } ], "protocol": { "version": "NMDRHT-response-models/1", "seed": 20261013, "frozenBeforeNewTraining": true, "M4": { "modelId": "m4-measured-decay-v1", "target": "Published transcript L2FC_kdeg_tx (EZbakR log2 change in degradation rate, N-AID-UPF1 12h IAA); retain sign, not an annotation-class probability", "mapping": "Exact NMDRHT transcript ID to native NMDRHT.v1.2 GTF. Validate gene, chromosome, strand, bounds, exon count and spliced length. Use source geometry/composition inputs only; refuse non-coding/inconsistent coding geometry. No outcome-derived NMD labels, relevance, expression, ribosome occupancy, or assay-response predictors." }, "M5": { "modelId": "m5-measured-te-response-v1", "target": "Published gene L2FC_TE (log2 translation-efficiency response to UPF1 depletion), not basal TE", "mapping": "Exact GENCODE.v42 protein-coding transcript FASTA to exact gene ID; deterministic representative longest CDS then transcript ID, selected without outcome. Full source sequence plus source-declared UTR/CDS geometry for shared sequence-context feature extraction. No RNA/RPF response predictors because they algebraically define TE." }, "holdout": "All eligible genes located on chr8; never used in fitting or configuration selection. Every transcript of a test gene stays in test. All other genes split by SHA256('nmd-response-val:'+stableGeneID), 80% train/20% validation. Exclude any cross-chromosome gene identity ambiguity. Previous five displayed NMDRHT audit-example genes cannot enter test.", "qualification": "Input mapping is independent of outcomes. Missing/nonfinite targets excluded only after frozen input partition membership; test targets are accessed by evaluator after every prediction is frozen. Do not filter test by target significance or response magnitude.", "fits": "Exactly six validation-selected configurations per family on identical training/validation rows: LightGBM 300 trees leaves 15/31/63 x L2 1/10; ExtraTrees 300 trees minleaf2/5/10 x max_features .5/1; scaled Ridge alpha .1/1/10/100/1000/10000. Median baseline plus M4 fitted 50nt-only rule comparator. No post-selection refit or test tuning.", "primaryMetric": "Spearman; report MAE and R2 as secondary diagnostics. Gene-weighted training so M4 genes with many isoforms cannot dominate. Paired whole-gene bootstrap 1000 rounds; Bonferroni-adjusted intervals within model comparison family.", "preserve": "New model/evidence identities only. Experimental response routes/results do not replace prior M4 annotation classifier or M5 basal-TE models.", "limitations": [ "Fresh gene/chromosome holdout for new measured-response endpoints within the same published study, not an independent experimental study", "Published workbook was previously inspected for schema, counts and five example rows; not represented as a newly collected or never-downloaded dataset", "Previous annotation/basal-TE models may have seen gene inputs; their weights are not used to fit these response models", "No therapeutic rescue or patentability inference; negative baseline comparisons remain visible" ] }, "qualification": { "sourceAssembly": "GRCh38 / native NMDRHT.v1.2 for M4; GENCODE.v42 exact FASTA/GTF for M5", "responseInputsForbidden": [ "NMD relevance/status/bin", "expression/RNA response", "RPF or TE response", "degradation response or adjusted p-value" ], "studyIndependence": "Gene/chromosome holdout within one published experiment; no independent-lab or independent-study confirmation", "ascertainment": "Native NMDRHT assembly uses control/depleted RNA samples; chromosome holdout does not remove assay or transcript-discovery ascertainment effects", "finiteMaskAudit": "M5 input partitions were frozen before response processing. A global finite-response availability mask was computed during development feature preparation, but an explicit test-inclusion branch retained every reserved test gene, including all 474 missing-target rows. Test numerical responses were never supplied to fitting or selection. This parsing is disclosed; not represented as never-parsed outcome bytes.", "decayContrast": "The author EZbakR code compares the N-AID-UPF1 12-hour state with control at 0 hours. This published contrast can include time/background effects; it is not an isolated, time-matched causal NMD effect." }, "defaultModelsReplaced": false, "overallSuperiorityEstablished": false }