遇见数据集

Updated DLDN-Bench De Novo Sequencing Benchmark Outputs with InstaNovo v1.2.2

收藏
Zenodo2026-06-19 更新2026-06-21 收录
官方服务:

资源简介:

This record contains benchmark outputs for an updated DLDN-Bench de novo sequencing benchmarking analysis that adds InstaNovo v1.2.2 to the published preprint benchmark. Predictions were generated with the InstaNovo v1.2.2 Python package using the instanovo-v1.2.0 transformer checkpoint, and were compared against the published Zenodo InstaNovo v1.1 predictions and the other tool outputs used in the original DLDN benchmark. The benchmark covers four PRIDE datasets: PXD043425, PXD006882, PXD012824, and PXD043200. The uploaded files were generated with the DLDN-Bench reproduction workflow. Included files: Four InstaNovo v1.2 prediction CSV files, one per PRIDE dataset. Reproduced and updated Figure 1 AUC summary tables in CSV, JSON, and Markdown format. Peptide-level and amino-acid-level precision-coverage plot PNGs for the original published all-tool benchmark and the updated benchmark including InstaNovo v1.2. True-positive overlap Venn diagram PNGs for the original published all-tool benchmark and the updated benchmark including InstaNovo v1.2. Workflow logs and the Zenodo download manifest used by the analysis. The large intermediate precision-coverage curve table CSVs are intentionally not included in this upload. These tables can be regenerated from the prediction files and DLDN-Bench analysis scripts, and can also be provided on request. Files are uploaded as individual Zenodo assets. Where a file originated from a subdirectory in the analysis output, the path separators are represented by double underscores in the uploaded filename. For the PXD006882 Figure 1 benchmark rowset corresponding to the quoted DLDN preprint values, the reproduced original peptide-level AUCs were 0.997318 for MS-GF+, 0.981222 for ContraNovo, and 0.973898 for InstaNovo v1.1. The updated rowset including InstaNovo v1.2 gave peptide-level AUCs of 0.997395 for MS-GF+, 0.982482 for ContraNovo, 0.975493 for InstaNovo v1.1, and 0.985262 for InstaNovo v1.2. At the amino-acid level for the same PXD006882 benchmark, the reproduced original AUCs were 0.999165 for MS-GF+, 0.990497 for ContraNovo, and 0.984830 for InstaNovo v1.1. The updated rowset including InstaNovo v1.2 gave amino-acid AUCs of 0.999210 for MS-GF+, 0.991443 for ContraNovo, 0.986267 for InstaNovo v1.1, and 0.995177 for InstaNovo v1.2. Across all four datasets in the aggregate updated rowset, InstaNovo v1.2 reached a peptide-level AUC of 0.957412 and an amino-acid-level AUC of 0.981321, compared with ContraNovo at 0.950945 and 0.972711 respectively. The corresponding source code and workflow scripts are maintained in the aichor branch of the BioGeek/DLDN-Bench GitHub repository, which is a fork of the main DLDN-Bench repository.

提供机构:
Zenodo
创建时间:
2026-06-19
二维码
社区交流群
二维码
科研交流群
商业服务