2026-10-07T19:29:50Z https://api.figshare.com/v2/oai

oai:figshare.com:article/30147601 2026-10-06T17:09:35Z category_24184 category_24196 item_type_3 month_year_10_2026

Analysis outputs Jemma Daniel (22263574) Bioinformatic methods development Proteomics and metabolomics De Novo Peptide Sequencing False Discovery Rate Estimation Procedure Proteomics Peptide identificatoin Bottom-up mass spectrometry Peptide filtering Updated winnow analysis outputs (tabular results).Models and training data (Hugging Face, pinned revisions):<ul><li><a href="https://huggingface.co/datasets/InstaDeepAI/winnow-ms-datasets">InstaDeepAI/winnow-ms-datasets</a> @ 659802319d618a359de5ab90ec6b0195681e94a6 (training/evaluation spectra and InstaNovo predictions; DOI 10.57967/hf/6610)</li><li><a href="https://huggingface.co/InstaDeepAI/winnow-general-model">InstaDeepAI/winnow-general-model</a> @ e2089330dd59adb9685e5b3d7d61f0cd69a3bbb0 (calibrator for general_results/, feature_importance/, and related analyses; config key train_extra_small_mass_error_da; DOI 10.57967/hf/6611)</li><li><a href="https://huggingface.co/InstaDeepAI/winnow-helaqc-model">InstaDeepAI/winnow-helaqc-model</a> @ d56542b961eac7d896e51bf0716a242fc394ab1f (InstaNovo HeLa calibrator (helaqc_results/instanovo/); DOI 10.57967/hf/6612)</li></ul>Additional calibrators (Figshare, same project/collection):<ul><li>Additional HeLa Single Shot Models (<a href="https://doi.org/10.6084/m9.figshare.32744946.v2" target="_blank" rel="noreferrer">10.6084/m9.figshare.32744946.v2</a>) (helaqc_results/casanovo/ and helaqc_results/primenovo/) Folders: casanovo_helaqc/, primenovo_helaqc/</li><li>Hold-one-out generalisation models (<a href="https://doi.org/10.6084/m9.figshare.30147364.v3" target="_blank" rel="noreferrer">https://doi.org/10.6084/m9.figshare.30147364.v3</a>) (companion to generalisation/calibrator_generalisation_results.csv) Folders: trained_on_gluc/, trained_on_hepg2/, trained_on_helaqc/, trained_on_herceptin/, trained_on_immuno/, trained_on_sbrodae/, trained_on_snakevenoms/, trained_on_tplantibodies/, trained_on_woundfluids/, ood_feature_cache/</li></ul>Outputs are organised into folders:<ul><li>ablations/ — Feature-ablation evaluation: aggregated metrics across withheld feature groups (ablation_summary.csv).</li><li>fdr_overlap/ — Winnow vs database-search overlap at 1 %, 5 %, and 10 % nominal FDR: retained PSM and unique-peptide counts, discordance categories, and per-project overlap summaries.</li><li>feature_importance/ — Feature-importance analysis for the general model on PXD014877 (C. elegans): permutation importance (perm_importance.pkl) and SHAP values (shap_values.pkl).</li><li>general_results/ — General-model evaluation on external benchmark datasets from InstaDeepAI/winnow-ms-datasets (general_model_evaluation/). labelled/ holds database-search reference runs; full/ holds full-search predictions. Each project folder contains metadata.csv and preds_and_fdr_metrics.csv.</li><li>generalisation/ — Leave-one-source-out calibrator generalisation metrics (calibrator_generalisation_results_slim.parquet). Corresponding calibrator checkpoints are in Figshare article 30147364 (trained_on_*/ folders, one model per held-out training source).</li><li>helaqc_results/ — HeLa QC benchmark (PXD044934): InstaNovo, Casanovo, and PrimeNovo prediction outputs. instanovo/ uses InstaDeepAI/winnow-helaqc-model; casanovo/ and primenovo/ use the HeLa calibrators in Figshare article 32744946 (casanovo_helaqc/, primenovo_helaqc/). Layout: {tool}/{split}/metadata.csv and preds_and_fdr_metrics.csv, where split is test (held-out labelled spectra), unlabelled only, or full search space less the training set.</li><li>novelty/ — Novel-peptide and non-tryptic digest analyses: calibration behaviour on peptides outside the standard tryptic training distribution (summary and per-dataset CSV tables).</li><li>upscored_fps/ — Up-scored false positives: false positives pushed into high-confidence regions by calibration, compared with true positives (upscored_summary.csv, upscored_fp_detail.csv).</li><li>external_peptide_holdout_benchmark/ — aggregated FDR / discovery metrics for Winnow, NovoBoard, and Glissade on HeLa QC and C. elegans in an external peptide-level FDR benchmark (external_peptide_holdout_results.csv, external_peptide_holdout_acceptance.csv, external_peptide_holdout_error_gain.csv).</li><li>fdr_tool_comparison/ — aggregated FDR / discovery metrics for Winnow and NovoBoard on HeLa QC and C. elegans in a PSM-level FDR benchmark (fdr_method_comparison_curves.csv, fdr_method_comparison_acceptance.csv, fdr_method_comparison_error_gain.csv)</li><li>fdr_benchmark_inputs/ — inputs needed to rerun the PSM-level FDR comparison (plot_fdr_method_comparison.py) and the peptide-level external score-mixture benchmark (run_external_peptide_holdout_benchmark.py) on HeLa Single Shot and C. elegans. Layout: winnow_results/instanovo_{helaqc,celegans}_predictions_{test,unlabelled}/ (preds_and_fdr_metrics.csv plus slim metadata.csv with spectrum_id and confidence); novoboard/{helaqc,celegans}/novoboard/ (target and decoy CSVs at the organism-tuned decoy rates 0.50 / 0.70); novoboard/helaqc/{annotated_test,raw_unlabelled}.mgf (Scan twin pairing); models/instanovo_{helaqc,celegans}/metadata_train.parquet (slim columns for Glissade’s training matched reference; holdout script only). Reference proteomes are in Hugging Face InstaDeepAI/winnow-ms-datasets (fasta/).</li></ul>Predict outputs use paired metadata.csv (spectrum metadata) and preds_and_fdr_metrics.csv (per-candidate scores, calibration, and FDR/q-value columns). Column definitions: winnow docs/cli.md (predict output section). 2025-09-29T14:55:21Z Dataset Dataset 10.6084/m9.figshare.30147601.v9 https://figshare.com/articles/dataset/Analysis_outputs/30147601 CC BY 4.0