Galgo

Galgo identifies prognostic gene expression signatures in cancer using a bi-objective evolutionary meta-heuristic that optimizes gene signature cohesiveness and association with patient survival.

Key Features:

• Bi-objective evolutionary meta-heuristic: Simultaneously optimizes for gene signature cohesiveness and association with patient survival.
• Transcriptomic classifiers: Generates robust transcriptomic classifiers that are strongly associated with patient prognosis.
• Statistical and machine-learning analysis: Performs statistical and machine-learning analysis of tumor transcriptomic profiles.
• Meta-analysis across cohorts: Applied a meta-analytic approach to 35 large population-based transcriptomic biobanks encompassing four cancer types.
• Comparative performance: Produced signatures that were superior predictors of patient survival in colorectal cancer and lung adenocarcinoma compared to existing molecular classification schemes.
• Breast cancer benchmarking: Identified a breast cancer signature that outperformed PAM50, AIMS, SCMGENE, and IntClust.
• Ovarian cancer benchmarking: Produced signatures with predictive power comparable to a consensus classification in high-grade serous ovarian cancer.
• Biological relevance: Generates partitions enriched for gene sets associated with the hallmarks of each disease.

Scientific Applications:

• Prognostic signature discovery: Identification of prognostic gene expression signatures across multiple cancer types.
• Cross-cohort meta-analysis: Integration and analysis of transcriptomic data across 35 population-based biobanks.
• Benchmarking molecular classifiers: Comparative evaluation against established schemes such as PAM50, AIMS, SCMGENE, and IntClust.
• Subtype biology investigation: Discovery of subtype-associated gene set enrichments linked to disease hallmarks.

Methodology:

Uses a bi-objective evolutionary meta-heuristic that simultaneously optimizes gene signature cohesiveness and patient survival, combined with statistical and machine-learning analysis of tumor transcriptomic profiles and a meta-analytic approach applied to 35 population-based transcriptomic biobanks across four cancer types.