{
  "id": 7454633,
  "title": "LucaCell: a sequence-centric foundation model for cross-species single-cell analysis",
  "url": "https://urgent.news/2026/09/14/lucacell-a-sequence-centric-foundation-model-for-cross-species-single",
  "topic": "science",
  "section": "Science",
  "published": "2026-09-14T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.09.08.750024v1?rss=1"
  },
  "original_language": "en",
  "account": "The recent development of LucaCell, a sequence-centric foundation model for single-cell analysis, marks a significant advancement in transcriptomic research. By representing genes through pre-trained mRNA sequence embeddings instead of static gene annotations, LucaCell transcends the limitations of traditional models that are confined to fixed gene identifiers. This innovative approach allows for seamless transfer of knowledge across various species and data types.\n\nLucaCell was pre-trained using an extensive dataset comprising 85 million human and mouse single cells. This broad training data set enables the model to generate accurate representations of cell types and gene expressions, even when cross-species analysis is required. To validate its effectiveness, LucaCell was evaluated across multiple domains including human and mouse gene expression profiles, human chromatin accessibility data, unaligned reads from over 50 prokaryotic taxa, and five influenza A virus genomes.\n\nOne of the most impressive capabilities of LucaCell is its ability to facilitate manual-mapping-free cross-species cell type annotation. This feature is particularly significant as it eliminates the need for time-consuming manual mapping, thereby streamlining the analysis process. Furthermore, LucaCell's unique alignment-free microbial embedding framework is capable of distinguishing between bacterial species identity and intra-species physiological states. This dual capability provides a comprehensive understanding of microbial communities, which is invaluable in fields such as microbiome research.\n\nIn addition to its cross-species capabilities, LucaCell also enhances gene expression reconstruction. By incorporating donor-specific exonic SNP information into mRNA sequence embeddings, the model achieves more precise gene expression profiles. This improved gene expression reconstruction is a critical step towards understanding complex biological processes and pathways.\n\nLastly, LucaCell's predictive capabilities are demonstrated through its ability to predict cellular viral load across different influenza A virus strains. This predictive power extends to identifying infection-like transcriptional states in mock-infected cells, providing insights into the early stages of viral infections. These findings collectively underscore the potential of sequence-informed gene representation in enhancing the generalization of single-cell foundation models.",
  "summary": "Single-cell foundation models have transformed transcriptomic analysis, yet most rely on fixed gene identifiers that limit transfer across species and data types. Here we present LucaCell, a sequence-centric foundation model that represents genes through pre-trained mRNA sequence embeddings rather than static gene annotations. Gene expression is discretized into bins and modeled with a…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}