PALACE method improves phage genome assembly from metagenomes using conjugate graph and deep learning (nature.com)

• PALACE integrates homology-based and deep-learning methods in a conjugate-graph framework to assemble phage genomes from metagenomes.
• Achieved F1 scores of 0.92–1.00 on simulated data, outperforming current best methods by significant margins.
• Applied to 914 human gut metagenomes, yielded 5,306 high-quality phage genomes with 55.98% improvement in median completeness over next best method.
• Phages from colorectal cancer patients showed enrichment of metabolic factors, suggesting adaptation to tumor microenvironment.

"Researchers at City University of Hong Kong developed PALACE, a conjugate-graph-based framework that integrates homology and deep learning to assemble high-quality phage genomes from metagenomic data. On simulated data, PALACE achieves F1 scores of 0.92–1.00, outperforming the second-best method by 0.21–0.48. Applied to 914 human gut metagenomes, it produced 5,306 high-quality phage genomes, improving median completeness by 55.98% over the next best benchmark. Phages from colorectal cancer patients showed enrichment of metabolic factors, suggesting adaptation to the tumor environment."