Microbial sequences and metagenomic classification. Explore models, applicable tests and evaluation requirements.
These are candidates, not a ranked list. An applicable test may need a trained head, a scoring adapter, a specialist comparison, or access to a hosted service. No result is implied by appearing here.
Public checkpoints; official local inference needs CUDA hardware and substantial memory.
Official source ↗Paper-reported results →Genome likelihood or embedding output needs promoter calibration.
Public model family and mini checkpoint.
Official source ↗Paper-reported results →Public promoter examples use task-specific fitting.
Sequence classifier needs training on a declared taxonomic split.
Report performance on clades excluded from head training.
Public Apache 2.0 checkpoint; 512-token context and large local memory requirement.
Official source ↗Paper-reported results →Metagenomic sequence embeddings need a read-classification head.
Fit a head on known taxa and test taxonomic shift.
Public classifier; database build/version must be pinned separately.
Official source ↗Paper-reported results →Remove held-out clades from the reference database; score retained ancestor ranks or unclassified reads, not exact labels absent from the database.
Native read classification with a pinned reference database.
Public profiler; marker database version must be pinned separately.
Official source ↗Paper-reported results →Each candidate test needs a defined dataset, split, metric and run protocol before it can become a benchmark.
Classify promoter activity from microbial DNA sequence.
Classify held-out phage or pathogen sequences and record taxonomic distance.
Hold clades out of downstream fitting and reference databases; evaluate known ancestor labels or unknown-taxon detection, and audit pretraining overlap separately.
Estimate taxon abundances in metagenomic samples.
Model sources and access descriptions checked 2026-09-16. Confirm licenses, terms and checkpoint revisions before any run.