Long Read RNA Sequencing in Primary Lung Cell Types Reveals Principles of Nonsense-Mediated Decay

Source data for: long-read RNA sequencing in primary lung cell types reveals principles of nonsense-mediated decay This record contains the starting data for every analysis in the paper: the processed quantification, annotation and model outputs that the analysis code reads but cannot regenerate. Nothing here is an intermediate the code also writes — those are rebuilt from these files, by design. Design. Paired short-read and PacBio Iso-Seq long-read RNA sequencing of four primary human lung cell types — alveolar type 2 (AT2), large airway epithelial (LAE), fibroblast (FB) and microvascular endothelial (MV) — under SMG1 inhibition (SMG1i) versus vehicle (DMSO). 26 samples: 13 donor-matched DMSO/SMG1i pairs. Cell-type labels are the published ones throughout (AT2, LAE, FB, MV). Some analysis code uses the internal codes AT and DD for AT2 and LAE respectively and maps them at load time. How the record is packaged Twelve files, 642 MB. Zenodo lists files flat, so the four directories ship as archives; extract them in place and the original layout is restored, which is what the analysis code expects. unzip sqanti.zip && unzip annotation.zip && unzip model.zip && unzip tan_2025_supplementary.zip Split by directory rather than bundled, so partial download works: the model exports are 6.7 MB on their own, and the Tan et al. tables are needed only for the Section 2 comparative reanalysis. UPLOAD_MANIFEST.sha256 checksums the download; each archive also carries an internal manifest for its extracted contents. 4.3 GB extracted — the SQANTI3 text compresses about 12×, while the Tan et al. .xlsx files are already compressed and gain almost nothing. Contents Quantification — the primary inputs file size shape what it is nmd_isocall_counts_4ct.csv 47 MB 614,993 × 26 Long-read isoform counts from the Isocall pipeline, one column per sample. The unfiltered isoform universe; the analysis applies its own expression filter (see below). nmd_lungcells_counts_4ct.csv 47 MB 614,992 × 26 Long-read isoform counts over the SQANTI3-filtered isoform set. salmon_gene_counts_4ct.csv 13 MB 46,571 × 26 Short-read gene-level counts (Salmon, length-scaled TPM→counts, via nf-core/rnaseq). pheno_4ct.csv 4 KB 26 × 5 Sample metadata: sample_name, donor, treatment, cell_type, bam. The join key for every count matrix. Scope is native, not filtered downstream. These matrices contain exactly the 26 samples of the published analysis. Two exclusions applied upstream are already reflected here: donor 029T (a PCA outlier, both treatment partners removed to keep the design balanced) and the air–liquid interface and small-airway conditions, which are outside the paper's scope. Code that consumes these files asserts the scope rather than re-filtering it, so an unexpected sample fails loudly. Transcript annotation file size what it is nmd_isocall_4ct.gtf.gz 41 MB Isoform models for the long-read universe; source of the transcript→gene map and exonic structure. sqanti.zip 332 MB (4.0 GB extracted) SQANTI3 output for the long-read transcriptome — structural classification, corrected models, and sequence. Treated as a starting point: the analysis begins from these calls and does not re-run SQANTI3. annotation.zip 6.9 MB Ensembl v115 gene- and transcript-level identifier maps (gmap_ENSGv115_*.rds, gmap_txlevel_ENSGv115_*.rds) used for symbol and biotype resolution; ensembl115_human_paralogs.tsv.gz with its README, the within-species paralog table behind the train/test leakage screen — homology is not derivable from a GTF, so it is deposited rather than recomputed; srsf8_ENST00000638426_reference.json, one reference transcript captured from the Ensembl REST API so the RBP section runs without network access; and msigdb_hallmark_H.tsv, the MSigDB HALLMARK collection (50 gene sets, 7,322 gene-set/symbol pairs, db_version 2026.1.Hs) captured so the productive-response section resolves hallmark membership offline instead of downloading MSigDB at run time. Key files inside sqanti.zip: nmd_lungcells_classification.txt (structural category per isoform), nmd_lungcells_filtered.gtf (filtered models), nmd_lungcells_corrected.cds.gff3 (CDS), nmd_lungcells_corrected.fasta (sequence). Deep-learning model — weights, predictions and interpretability model.zip, 6.7M, 28 files extracted. The model predicts NMD susceptibility from sequence and ORF structure. Training requires a GPU cluster, so the trained weights and all derived outputs are deposited rather than expecting a reader to retrain. Filenames are the producing run's own and are not shortened: they carry the window configuration, the member seed and the evaluation split, because those are what distinguish one model's outputs from another's. file(s) what it is best_model_atg1000_stop1000_seed42.pt Trained weights (1000 nt windows around the start and stop codons, member seed 42). metrics_atg1000_stop1000_seed42_test_clean.json Held-out performance: AUC 0.9257, AUPRC 0.8175 over the 10,522-isoform test_clean split (2,405 NMD susceptible). evaluation_class: final_test — evaluated once. predictions_atg1000_stop1000_seed42_all.tsv Per-isoform predictions for all 41,776 isoforms, with h5_split (train 26,720 / val 4,356 / test 10,522 / test_paralog 122 / val_paralog 56), label, logit, prob. predictions_atg1000_stop1000_seed42_test_clean.tsv The held-out split alone (10,522 rows) — a strict subset of the above. kernel_shap_branch_atg1000_stop1000_seed42_all.tsv KernelSHAP attributions per input branch (sequence, ORF structure, stop-window), all isoforms. Exact Shapley by enumeration over three players — not the sampling estimator, so shap.KernelExplainer will not reproduce it. deepshap_summary_atg1000_stop1000_seed42_{joint,structural,atg-stop}_run{1..5}.tsv DeepSHAP feature attributions: three DISTINCT decompositions, five replicates each. The mode is in the filename deliberately — they are different analyses and must not be pooled. shap_profile_{atg,stop}_joint_atg1000_stop1000_seed42.tsv Per-nucleotide signed SHAP around the start and stop codons. motif_logo_atg_atg1000_stop1000_seed42_run1.tsv, motif_logos_stop_atg1000_stop1000_seed42_run1.tsv Position-weight matrices for the sequence-logo panels. The two differ in stem (logo / logos) because the producers do; the names are not normalised here. uorf_attention_{metrics,predictions,subgroup_counts}.tsv, uorf_features_in_priority_slots.tsv Model attention over candidate upstream ORFs, and the subgroup counts behind the uORF figure panels. The model was trained on the universe these files rebuild — 41,776 isoforms. Splits are chromosome-deterministic and sequences are byte-identical to the deposited FASTA across the whole universe. The interpretability exports are deposited as starting points: the figures and statistics in the paper's model section are computed from these tables, so they can be reproduced without a GPU. Citations If these files are used please cite these publications: Path Contents gerstberger_2014_rbp_census.csv Census of human RNA-binding proteins — 1,542 RBPs with Ensembl gene id, consensus RNA target and domain annotation. Gerstberger S, Hafner M, Tuschl T, Nature Reviews Genetics 2014;15:829–845, doi:10.1038/nrg3813, Supplementary Table S3. encode_rbp_roster_vannostrand2020.csv ENCODE eCLIP RBP roster — 356 experimentally characterised RBPs with functional-class annotation. Van Nostrand EL et al., Nature 2020;583:711–719, doi:10.1038/s41586-020-2077-3. tan_2025_supplementary.zip Tan et al. 2025 Supplementary Tables S1, S2, S4 and S6 (.xlsx) — the transcript-level differential-expression tables read by the Section 2 comparative reanalysis. Tan K, Sebat J, Wilkinson MF, Nucleic Acids Research 2025;53, doi:10.1093/nar/gkaf395. The two RBP tables are unmodified apart from conversion to CSV. The four Tan et al. tables are the publisher's .xlsx files byte-for-byte, renamed only to normalise the filenames. The functional-class columns used in Section 2 come from the ENCODE roster, not from the Gerstberger census, whose columns are gene symbol, Ensembl gene id, consensus RNA target and domains only. Reproducing from this record Obtain the analysis code (see Code availability in the paper). Extract the three archives in place, then point the code's configuration at this directory as its source-data root. The pipeline rebuilds its own intermediates from the files here — expression objects (DGELists), the limma and mashr differential-expression results, and the isoform-pair sets. Nothing downstream of these files is deposited, by design. The per-cell-type mashr tables that the isoform-pair analyses consume, for instance, are written by the code from the count matrices and sample table above; they are outputs of the shipped pipeline rather than inputs to it. Reproduction has been verified for the data-preparation stage and is being extended through the remaining analyses; the code repository records the current state. Raw sequencing reads are not in this record. They are in GEO under accession GSE329233, which is the appropriate archive for them. This record holds the processed starting data. Notes for reusers The isoform universe is unfiltered. nmd_isocall_counts_4ct.csv carries 614,993 isoforms; the published analyses apply an expression filter (edgeR::filterByExpr) that reduces this to 162,800. If you compare against a published count, check which universe it refers to — several published numbers are stated over one and several over the other. predictions_atg1000_stop1000_seed42_test_clean.tsv is redundant with the test rows of predictions_atg1000_stop1000_seed42_all.tsv. Both are included because published values were computed from each. Cell-type label mapping. AT2 ≡ internal AT; LAE ≡ internal DD. Sample column names in the count matrices follow <donor>_<treatment>_<cellType>. sqanti.zip dominates the record — 332 MB of the 642 MB download, and 4.0 GB of the 4.3 GB extracted. tan_2025_supplementary.zip is the next largest at 147 MB and is needed only for the Section 2 comparative reanalysis. Everything else together is under 165 MB and is enough for the quantification, differential-expression and model analyses; sqanti.zip is needed for the isoform-structure and PTC work.