Biological models

318 model records in release 2026-10-23-5347695500ed (release manifest, gzip). Showing 217–240; page 10 of 14.

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

• Orthrus

  RNA and transcriptomes

  Mamba RNA encoder trained contrastively so that splice isoforms and orthologous transcripts have similar embeddings, used to predict mature mRNA properties.

• PAGTN

  Protein function

  Path-augmented graph transformer network.

• Pangolin

  DNA and genomes

  Pangolin predicts splice-site strength and changes caused by genetic variants.

• PGraphDTA

  Molecular interactions

  PGraphDTA: GraphDTA variant using protein language model embeddings (ESM-2 variant without contact maps used here).

• PhyloGPN

  DNA and genomes

  Phylogenetic genomic language model (PhyloGPN).

• PiFold

  Molecular interactions · Proteins and complexes

  PiFold graph-transformer inverse-folding model.

• PlantCaduceus

  DNA and genomes

  Caduceus and Mamba DNA language model pretrained on 16 angiosperm genomes, fine-tuned to predict translation start and stop sites and splice sites across plant species and to score deleterious variants.

• PlantRNA-FM

  RNA and transcriptomes

  PlantRNA-FM RNA foundation model pretrained on plant transcriptomes.

• PLM-GraphDTA

  Molecular interactions

  PLM-GraphDTA: GraphDTA drug encoders with frozen protein language model embeddings as the protein representation.

• PoET

  Proteins and complexes

  Autoregressive transformer that models whole protein families as sequences of sequences and scores or generates variants conditioned on homologous sequences, scored in ProteinGym as a zero-shot substitution baseline.

• PredFull

  Proteins and complexes

  PredFull residual CNN full-spectrum predictor with squeeze-excitation (about 8.2M parameters).

• ProGen2

  Proteins and complexes

  Suite of protein language models of up to 6.4 billion parameters trained on sequence datasets drawn from over a billion proteins, which generate sequences and predict fitness without fine-tuning; ProteinGym scores it as a zero-shot single-sequence substitution baseline.

• ProGen3

  Proteins and complexes

  Family of sparse generative protein language models of up to 46 billion parameters pretrained on 1.5 trillion amino-acid tokens for sequence generation and fitness prediction, scored in ProteinGym as a zero-shot single-sequence substitution baseline.

• ProkBERT

  Microbes and communities

  ProkBERT is a family of microbial DNA encoders used for sequence representation, promoter prediction and phage identification.

• PromoterAI

  DNA and genomes

  Deep learning model that predicts the impact of promoter variants on gene expression. Scores range from -1 to 1 (negative for under-expression, positive for over-expression); the README recommends thresholds of 0.1, 0.2 and 0.5 in absolute value. Supports single nucleotide variants and left-normalised indels.

• Prosit

  Proteins and complexes

  Prosit GRU encoder-decoder fragment intensity predictor (about 2.5M parameters).

• Prosit Transformer

  Proteins and complexes

  Prosit Transformer fragment intensity predictor (about 86M parameters).

• ProSST

  Proteins and complexes

  Transformer protein language model that combines the residue sequence with quantized structure tokens through disentangled attention, pretrained by masked language modelling, scored in ProteinGym as a zero-shot substitution baseline from sequence and structure.

• ProstT5

  Molecular interactions · Proteins and complexes

  ProstT5 bilingual sequence and 3Di structure protein language model.

• Prot-T5

  Protein function

  ProtT5 protein language model, as named in the resource-usage table.

• Prot-T5-XL

  Protein function

  ProtT5-XL protein language model.

• ProtBert

  Protein function

  ProtBert protein language model.

• ProteinMPNN

  Proteins and complexes

  ProteinMPNN designs amino-acid sequences for a supplied protein backbone.

  Record: catalog-model-proteinmpnn

• ProteinMPNN

  Protein structure

  ProteinMPNN designs amino-acid sequences for a supplied protein backbone.

  Record: discovery-model-proteinmpnn

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.