Uncovering cancer vulnerabilities by machine learning prediction of synthetic lethality.
Source check: Source checked. Not independent scientific confirmation.
Finding: PARIS inferred synthetic-lethal interactions from DepMap CRISPR and multi-omic data. Follow-up CRISPR assays supported an ALDH2-BRIP1 dependency in low-ALDH2 cancer-cell models; this is preclinical cell evidence, not patient stratification or treatment benefit.
Technology: PARIS learned synthetic-lethality predictor
Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.
- AI contribution
- PARIS uses random-forest importance scores and Boruta feature selection to explain CRISPR dependency scores from mutation and expression features across DepMap cell lines, then ranks candidate synthetic-lethal interactions.
- Test and result
- PARIS predicted 15 high-confidence interactions across 549 DNA-damage-repair genes. Experimental follow-up tested ALDH2-BRIP1 in breast cancer cell lines and engineered RPE-1 cells using CRISPR knockouts, colony formation and DNA-damage readouts. The paper separately reports experimental validation of CDKN2A-TYMP-TYMS.
- Limitation / still unknown
- Validation was in cell models and does not establish patient stratification or treatment benefit. The authors note heterozygous Fanconi-anemia-pathway mutations in several cancer lines as a possible background confounder.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://doi.org/10.1186/s12943-021-01405-8
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 18 Nov 2026. Review scope: Full PMC paper, including model description, experimental methods, results and discussion..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.