rewirebio.iobenchmarks

Biological models

Understand model architectures, biological inputs, access requirements and the evidence from linked evaluations.

374 model records in release 2026-11-21-c8bea25e5abb (release manifest, gzip). Showing 97–120; page 5 of 16.

Evaluated configurations, methods and pipelines are listed separately. Names alone do not establish equivalent models or checkpoints.

  • Enchant v3

    Molecular interactions

    Enchant v3 is Iambic's multimodal transformer for drug discovery. It predicts molecular properties and preclinical and clinical endpoints from structures, sequences, omics, assay data, text and images, and Iambic uses it to choose compounds and prioritise experiments.

  • EnCodon

    DNA and genomes

    Encoder-only codon language model.

  • Enformer

    DNA and genomes

    Convolution and Transformer model trained to predict genomic tracks from sequences of up to about 200 kb.

  • Enigma-Distilled

    DNA and genomes

    Distilled supervised regulatory sequence model named Enigma-Distilled in GUANinE v1.1.

  • ERNIE-RNA

    RNA and transcriptomes

    ERNIE-RNA RNA language model with structure-enhanced representations.

  • ESA-dna2vec

    DNA and genomes

    ESA-dna2vec DNA embedding model.

  • ESCOTT

    Proteins and complexes

    Missense effect predictor that ProteinGym lists with multiple sequence alignment and structure inputs under the PRESCOTT reference, which describes an epistatic and structural model of missense effects that also uses population allele frequencies; scored as a zero-shot substitution baseline.

  • ESM-1b

    Proteins and complexes

    Deep contextual protein language model trained without labels on 250 million protein sequences, whose learned representations support mutational-effect prediction; ProteinGym scores it as a zero-shot single-sequence substitution baseline.

  • ESM-1v

    Protein function

    ESM-1v provides protein language models intended for zero-shot variant-effect scoring.

  • ESM-2

    Protein function

    ESM-2 is a family of protein sequence encoders that produce representations for downstream protein analyses.

  • ESM-C

    Protein function

    ESM Cambrian protein language model family, evaluated at 300M and 600M parameters.

  • ESM-IF1

    Protein structure

    ESM-IF1 designs protein sequences conditioned on backbone coordinates.

  • ESM3

    Protein structure

    ESM3 generates and completes protein sequence, structure and functional annotations using a shared multimodal representation.

  • ESMC

    Protein function

    ESM C learns protein sequence representations for downstream analysis.

  • ESMFold

    Protein structure

    ESMFold predicts protein structures directly from amino-acid sequence using ESM-2 representations.

  • ESMFold2

    Protein structure

    ESMFold2 predicts biomolecular structures from protein, DNA, RNA and ligand inputs, optionally using protein alignments.

  • EVE

    Proteins and complexes

    Deep generative model of sequence variation across organisms that predicts the effects and pathogenicity of protein variants without labels; ProteinGym scores it as a zero-shot alignment-based substitution baseline.

  • Evo

    DNA and genomes

    StripedHyena DNA language models of 7 billion parameters (Evo 1 and Evo 1.5), trained at single-nucleotide resolution on prokaryotic and phage genomes with 8 kb or 131 kb contexts, for zero-shot function prediction and sequence generation.

  • Evo 2

    DNA and genomes

    Evo 2 models and generates DNA over long contexts at single-nucleotide resolution.

  • FastSpel

    Proteins and complexes

    FastSpel lightweight linear spectrum predictor (about 50K parameters), used as a lower-bound reference.

  • FLARE

    Metabolomics

    Contrastive spectrum-to-molecule retrieval model that follows MVP with a stronger spectra encoder using subformula peak annotation and a richer similarity; it needs the target's molecular formula.

  • FlowDock

    Molecular interactions

    FlowDock: conditional flow-matching docking and affinity model built on ESMFold and NeuralPLexer components.

  • FlowPacker

    Protein structure

    FlowPacker builds all-atom side chains onto a protein backbone for a given sequence.

  • FLOWR.ROOT

    Molecular interactions

    FLOWR.ROOT: SE(3)-equivariant flow-matching model with joint structure, affinity and confidence heads.

This index reflects a dated catalogue, not an exhaustive census. Source checking does not mean independent reproduction; compare results only under compatible protocols, datasets and metrics.