MetaScreener: a robust dual-mode framework for directional prioritization of actionable signatures through multi-dataset and multi-approach integration.

Reliably quantifying the directionality of signatures from noisy omics data remains challenging, hindering their widespread applications in biomedical research and drug development. To address this, we developed MetaScreener, a dual-mode framework (DiffMetaScreener and CorMetaScreener) that integrated multiple datasets and a large number of analysis pipelines (>4, 000) for robust quantification of activating and inhibiting signals in gene signatures. Supervised application of MetaScreener on 9 colorectal cancer (CRC) training datasets prioritized positive and negative Wnt signaling signatures, respectively. Robustness benchmarking against random perturbations, including extreme noises, revealed that MetaScreener outperformed conventional approaches. Independent validation on 22 CRC datasets and a massive-scale single-cell dataset further confirmed the accuracy of the prioritized signatures. In the case study, parthenolide was discovered and experimentally validated as a dual inhibitor targeting both Wnt signaling and CRC metastasis. Collectively, MetaScreener demonstrates as a reliable and powerful tool for directional prioritization of both functional and actionable gene signatures.