Screening of Biosecurity Features in Metagenomic Data with Evo 2 Probes

Genomic foundation models such as Evo 2 learn rich sequence representations, but their value for biosecurity screening is largely unexplored. This study trains minimal linear and attention probes on frozen Evo 2 layer-26 activations to assess how much biosecurity-relevant signal is linearly accessible without fine-tuning the base model. Across held-out metagenomic test sets, the probes detect antimicrobial resistance (AMR) with strong discrimination: a linear probe achieves region-level ROC-AUC of 0.888 (mean-pool), rising to 0.977 with a single-head attention probe. The probes also resolve finer-grained AMR drug-class subcategories and separate them from unrelated functional genes. Bacterial virulence is decodable but weaker (region-level ROC-AUC 0.833). On simulated short reads, the AMR probe retains comparable ranking performance (read-level ROC-AUC 0.898), enabling evaluation before assembly in cost-prohibitive or unreliable settings. A complementary sparse-autoencoder analysis recovers interpretable resistance-associated features but proves less consistent than supervised probes. Together, these results position lightweight embedding-based probes as a fast, inexpensive first-pass detection layer for metagenomic biosurveillance.