Multi-omics integration and artificial intelligence for the conservation and utilization of local chicken genetic resources.

Local chicken genetic resources (LCGRs), representing a unique gene pool shaped by millennia of natural and artificial selection, not only sustain the supply of high-quality protein but also serve as useful biological models for deciphering the evolution of complex traits and environmental adaptability. However, extensive introgression from commercial breeds is causing rapid genetic erosion. Systems biology and multi-omics technologies are reshaping our understanding of the regulatory networks underlying local chicken genetic resources. This review synthesizes four advances. First, at the genomic reference level, long-read sequencing is driving a transition from single linear reference assemblies to high-quality, near-telomere-to-telomere assemblies and graph pangenomes, enabling the unbiased capture of structural variations and microchromosomes. Second, multi-omics studies have begun to integrate association-based evidence across multiple biological layers, including epigenetic variation, single-cell and spatial heterogeneity, cross-tissue metabolic relationships, and host-microbiome interactions. Third, commercial introgression is severe but strongly breed-dependent, affecting 0.64% to 21.52% of the genome across eight Chinese indigenous breeds. We examine how omics findings could be translated into conservation practice by integrating three components into a proposed closed-loop framework: dynamic early-warning monitoring based on effective population size, management of functional variants using a weighted genomic relationship matrix, and primordial germ cell cryopreservation and editing. These components sit at very different levels of evidence, and the complete pipeline has not yet been evaluated longitudinally in any conservation flock. Finally, in the realm of intelligent prediction, the mechanistic attribution provided by explainable artificial intelligence and the zero-shot variant effect prediction capabilities of cross-species genomic foundation models offer two complementary routes, neither of which has yet been applied to a local chicken population. Together, these advances are shifting local chicken genetic resources management from observation-based description toward mechanism-informed decision-making. Rather than reporting an accomplished transition, this review sets out an emerging and feasible roadmap and identifies the evidence gaps that must be closed before it can be implemented.