Large-scale automated machine reading discovers new cancer-driving mechanisms.
Level 5 - mechanism / opinion, no new human data
Computational methodology and bench informatics study without clinical trial data (graded by design analogy).
PubMed 30256986 · doi:10.1093/database/bay098
What was done
The authors developed Reach, an automated large-scale machine-reading system designed to extract mechanistic descriptions of biological processes from biomedical literature. They integrated the extracted pathway fragments with curated biological models to analyze mutually exclusive altered signaling pathways across seven distinct cancer types.
What was found
The authors reported that Reach extracts mechanistic descriptions with relatively high precision at high throughput. When combined with existing algorithms and curated models, the system identified and explained previously unidentified mutually exclusive altered signaling pathways in seven cancer types. The abstract provides no specific quantitative performance metrics, counts of papers processed, or numerical pathway discovery totals.
Why it matters
Biomedical publication volume far exceeds human manual curation capacity. Automated text-mining frameworks capable of causal extraction can help systematically map complex molecular mechanisms and generate testable hypotheses in cancer biology.
Limits
The abstract reports zero numerical metrics for extraction performance (such as precision, recall, or F1 scores) and does not quantify the specific findings. The results represent computationally inferred pathways rather than experimentally validated mechanisms.
Cited by
- supports Approximately one million new publications are indexed in peer-reviewed journals on PubMed each year.