Salmonella Transcriptional Regulatory Networks Reconstructed from 2,278 Articles

An ensemble of fine-tuned GPT models was used to reconstruct bacterial transcriptional regulatory networks from scientific literature. The method was evaluated on 264 full-text articles for Salmonella Typhimurium and then applied to a corpus of 2,278 articles. This dataset, a 160.5 KB PDF authored by José Romero-Vilchis and updated in April 2026, describes the approach and results.

Use Cases

• Training models for literature-based regulatory network prediction based on the described fine-tuning strategy.
• Benchmarking NLP methods for biological relation extraction based on the reported F1-scores and Matthews correlation coefficients.
• Analyzing Salmonella Typhimurium transcriptional networks based on the reconstructed network from 2,278 articles.
• Studying the link between transcriptional regulation and antimicrobial resistance based on the network's described alignment with biological knowledge.

Strengths

• Model performance is quantified with specific metrics (F1-Score > 0.87, Matthews correlation coefficient > 0.82).
• The final network reconstruction achieved a recall of 0.80 from a curated test set.
• The analysis is based on a substantial literature corpus of 2,278 articles.

Limitations

• The dataset is a 160.5 KB PDF, which is a very small file likely containing only a description of the method and results, not the underlying network data.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

figshare

Collection Method

Models fine-tuned on Escherichia coli literature and applied to Salmonella Typhimurium articles.

Time Range

Literature available at the time of the study.

Freshness

Last updated 2026-04-13 05:48:24; freshness should be verified.

Geography

null