TaxoFormer: New AI Model Predicts Protein Taxonomic Classification from Sequences
Researchers have introduced TaxoFormer, a hierarchical transformer model that predicts the complete taxonomic lineage of a protein from its amino acid sequence alone. The model encodes the entire NCBI phylogenetic tree—over 1.3 million nodes—into a vocabulary of just 15,000 tokens, then couples a pre-trained protein language model (ESM-2) with an autoregressive decoder trained on 188 million proteins. The approach offers a scalable, alignment-free alternative to traditional sequence-alignment methods for taxonomic annotation and suggests that explicitly structuring the output space can help neural networks learn biologically meaningful representations.
TaxoFormer addresses a fundamental machine learning challenge: predicting labels in extremely large, hierarchically organized output spaces. The system's central innovation is a structured tokenization scheme that losslessly compresses the NCBI phylogenetic tree—a graph exceeding 1.3 million nodes—into a compact 15,000-token vocabulary, making the full taxonomic hierarchy tractable for a generative model. By pairing this tokenization with ESM-2, a widely used pre-trained protein language model, and an autoregressive decoder trained with a standard cross-entropy loss, the authors test whether a simple generative objective is sufficient to capture complex biological structure. Trained on 188 million protein sequences, TaxoFormer achieves accurate lineage prediction across multiple taxonomic ranks without requiring sequence alignment, which is computationally expensive and can fail for highly divergent sequences. Beyond classification accuracy, the model implicitly learns a continuous latent space that reflects phylogenetic relationships, suggesting the representations are biologically grounded rather than superficial. The work positions itself as both a practical tool for large-scale metagenomic and genomic annotation and a methodological demonstration that output-space structure is a powerful inductive bias for representation learning.
What's missing
The preprint does not report benchmark comparisons against established alignment-based taxonomic classifiers (e.g., Kraken2, DIAMOND, or MMseqs2) in terms of speed and accuracy trade-offs, making it difficult to assess practical gains over existing tools. Generalization to highly novel or synthetic sequences outside the training distribution is not characterized, nor are failure modes at shallow versus deep taxonomic ranks quantified in detail. As a preprint, the work has not yet undergone peer review.
What different sources said
- bioRxivCenter
TaxoFormer: Hierarchical Transformer for Predicting the Full Taxonomic Lineage of Protein Sequences
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.