EvidenceNet: Source-Grounded Biomedical Evidence for HCC, CRC, and SLE

EvidenceNet is a collection of 18,755 source-grounded biomedical evidence records extracted from full-text literature, created by Chang Zong and released on figshare in May 2026. It focuses on three diseases: hepatocellular carcinoma (7,872 records), colorectal cancer (6,622 records), and systemic lupus erythematosus (4,261 records). The release includes companion graph files with up to 10,328 nodes and 49,756 edges, along with schemas, validation outputs, and extraction prompts.

Use Cases

• Train or evaluate biomedical relation extraction models based on the structured evidence-relation annotations.
• Build or analyze disease-specific knowledge graphs based on the provided node and edge files.
• Study evidence provenance and context in scientific literature based on the source-text and article-provenance fields.
• Benchmark automated evidence aggregation workflows using the provided validation and prompt templates.

Strengths

• Contains 18,755 total evidence records across three specific diseases.
• Includes companion graph representations with up to 49,756 edges for network analysis.
• Provides a detailed schema, validation outputs, and the prompt templates used for extraction.

Limitations

• Column-level documentation is absent; field semantics must be inferred from the schema after download.
• The geographic and temporal coverage of the source literature is not specified.
• Row count for individual files is unknown, which may limit suitability assessment.

Provenance

Source

Full-text disease-focused literature, extracted via a released workflow.

Collection Method

Likely involves automated extraction and aggregation using prompt templates, followed by manual audit and validation.

Freshness

Last updated 2026-05-18 14:38:49; freshness should be verified.