Artificial intelligence and multi-omics integration for cardiac regeneration: from cell fate decoding to clinical translation

Heart failure remains among the leading causes of death worldwide, and its burden is increasing as populations age. The adult mammalian heart has almost no capacity to replace cardiomyocytes lost to injury: most cardiomyocytes leave the cell cycle, and the infarcted muscle is replaced by scar, so that function declines progressively. This constraint is not fixed. Neonatal mice regenerate myocardium for a brief window that closes within the first postnatal week, whereas adult zebrafish retain the capacity throughout life; the machinery for regeneration therefore exists, but is held in check. Cell-based therapy was expected to circumvent the problem, yet its clinical benefit has remained modest: transplanted cells survive and engraft poorly, their fate resists precise control, and the host microenvironment actively suppresses regeneration. These shortcomings are symptoms of a single underlying difficulty—scale. The regulatory logic of cardiac regeneration is distributed across transcriptional control, intercellular communication, spatial microenvironments and protein interactions, so that no omics layer examined in isolation, however high its resolution, can recover the integrated picture from high-dimensional, sparse and heterogeneous data. This review proposes an artificial intelligence (AI)–multi-omics integration framework for cardiac regeneration. We survey the methodological progress that becomes available when single-cell transcriptomics, spatial transcriptomics and proteomics are modelled jointly, and examine what these approaches can and cannot establish about cardiac lineage cell fate determination, about regeneration trajectory reconstruction, and about the failing-heart microenvironment. We then consider the translational pathways, and the limits of the evidence for each. The resulting evidence is uneven: the sections on the failing-heart microenvironment rest largely on direct cardiac data, whereas most of the methodological progress has been made outside the heart and transfers only under additional assumptions. One qualification is therefore stated at the outset: evidence for AI in cardiac regeneration is still scarce. The review is positioned as a methodological framework and a forward-looking perspective rather than a synthesis of established conclusions, and the evidence supporting each principal claim is graded, from direct experimental evidence in the heart to methodology.