Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-2220
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-23T18:35:08.756261Z
status: published
lens: g_space
sector: science
headline: Quantifying Data Reuse in Proteomics Using PRIDE Downloads Statistics and a Semi-Supervised LLM-Based Framework
dek: Understanding how scientific datasets are accessed and reused is essential for resource planning and impact assessment. Here we present the PRIDE Archive download tracking infrastructure and a comprehensive analysis of 159.3 million download records from the PRIDE proteomics database (2021-2025), spanning 35,528 datasets accessed from 235 locations. The infrastructure includes nf-downloadstats, a scalable Nextflow pipeline for processing download logs, and DeepLogBot, a machine-learning framework that classifies…
gain_title: A semi-supervised classifier using LLM-assisted labels separated bot, institutional hub and independent user traffic in proteomics download logs, reaching 92.2% bot accuracy on held-out data.
problem_title: (none)
trace_subject: (none)
gain_reading: A semi-supervised classifier using LLM-assisted labels separated bot, institutional hub and independent user traffic in proteomics download logs, reaching 92.2% bot accuracy on held-out data.
gain_evidence: achieving 92.2% bot classification accuracy on a held-out test set
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers built download-tracking infrastructure for the PRIDE proteomics archive and analysed 159.3 million download records from 2021 to 2025, covering 35,528 datasets. Their DeepLogBot framework used labels produced by two language models, Claude and Qwen3, to sort traffic into bots, institutional hubs and independent users.

Separating automated from individual access gives repository operators a cleaner picture of how scientific data are actually reused, which matters for planning resources and assessing impact. The abstract reports performance only on a held-out test set from this one repository, so how well the approach transfers elsewhere is not established.
limitation: 
tag: Evidence-backed gain
key_points: The analysis covered 159.3 million download records from 2021-2025 across 35,528 datasets. | Downloads came from 235 locations, and after bot removal from 214 countries/regions. | The pipeline includes nf-downloadstats, a Nextflow tool for processing download logs. | 249 institutional download hubs were identified once bot traffic was separated. | The top five countries accounted for over 54% of independent user downloads.
rundown: The work has two components: nf-downloadstats, a scalable Nextflow pipeline that processes download logs, and DeepLogBot, a machine-learning framework that classifies traffic into bots, institutional download hubs and independent user downloads.

DeepLogBot starts with heuristic seed selection and then applies multi-LLM annotation using Claude and Qwen3 to produce consensus labels. It reached 92.2% bot classification accuracy on a held-out test set.

After bot traffic was removed, reuse proved concentrated. The United States, United Kingdom, Germany, China and Canada together made up over 54% of independent user downloads. The authors say the findings can inform repository infrastructure planning.
sources:
- peer_reviewed | PROTEOMICS | https://doi.org/10.1002/pmic.70200 | 2026-10-21
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

77ccf04eb75734192ade5b0f1a378cd1df2262caf1e563ff5815a6b3e1a03cbf

previous

0000000000000000000000000000000000000000000000000000000000000000