A Transferable Genomic Language Model Framework for Fungal Gene Essentiality Prediction

Predicting biological function from genomic sequence remains a major challenge in computational and systems biology. Here, we tested whether the genomic language model Evo2, which encodes context-dependent DNA sequence patterns into embeddings, enables prediction of essential genes in fungi, a phenotype central to fungal biology and antifungal target discovery. We found that model performance was constrained not by the type or complexity of the downstream classifier, but by the biological information contained in Evo2 DNA embeddings. Specifically, the information recoverable from these embeddings progressively declined for biological features further downstream of DNA sequence, revealing a bottleneck for predicting higher-order cellular phenotypes. We alleviated this bottleneck by integrating Evo2 embeddings with two sequence-informed, system-level features: ortholog-based essentiality and protein-protein interactions. The multimodal framework demonstrated consistent performance both within and across three evolutionarily divergent yeasts ( Candida albicans , Saccharomyces cerevisiae , and Schizosaccharomyces pombe ), and its predictions were supported by published experimental evidence when transferred to the filamentous mold Aspergillus fumigatus . These results establish our framework as a transferrable tool for predicting essential genes across fungal genomes, including species with limited or no experimentally determined essentially data.

Essential genes in fungi are promising targets for antifungal drug development, yet experimental, genome-wide essentiality screening remains slow, resource intensive, and difficult to scale. As a result, comprehensive gene essentiality profiles exist for only a few model species, including Candida albicans , Saccharomyces cerevisiae , and Schizosaccharomyces pombe . In this work, we demonstrate that genomic language models can be leveraged through transfer learning to predict fungal essential genes directly from genomic sequence, either without training data from the target species or after fine-tuning on limited target species data. Our transfer learning approach offers a promising strategy to accelerate antifungal target discovery across understudied fungal pathogens and clinical isolates.