AI-Driven Comparative Genomics for Human Regenerative Gene Discovery Using Axolotl as a Model

Humans typically repair injuries through incomplete structural restoration and fibrosis, whereas axolotls can regenerate complex appendages. Whether this difference reflects loss of key regeneration genes or incomplete deployment of conserved programs remains unresolved. Here, we integrated public axolotl limb regeneration RNA-seq data with cross-species mapping and supervised machine learning to prioritize human-side counterparts of regeneration-associated genes. A total of 710 genes were consistently upregulated across 3, 6, and 14 days post amputation. OrthoFinder directly mapped 558 genes to human orthologs, and a local BLASTp rescue workflow increased the total number of genes with human protein matches to 690, leaving 20 unresolved. After removing genes already used in model training, 640 candidates were scored by a Random Forest classifier, and 590 exceeded the workflow decision threshold for downstream analysis. Functional enrichment with g:Profiler showed strong overrepresentation of cell cycle, DNA replication, embryo development, protein localization to chromosome, and nuclear organization terms. Together, these findings support a model in which humans retain a broad set of genes associated with regenerative programs observed in axolotl, while also emphasizing that computational conservation and prioritization alone do not prove functional activation in human injury contexts. This study provides a ranked candidate set for future experimental validation.