The Best 59 Protein Model Tools in 2025

ESM-2 is a next-generation protein model trained with masked language modeling objectives, suitable for fine-tuning on various downstream tasks with protein sequences as input.

ESM-2 is a next-generation protein model trained with masked language modeling objectives, suitable for fine-tuning on various protein sequence tasks.

ESM-2 is a state-of-the-art protein model trained on masked language modeling objectives, suitable for protein sequence analysis and prediction tasks

ESM-2 is a cutting-edge protein model trained on masked language modeling objectives, suitable for various protein sequence analysis tasks

A protein sequence pre-training model based on BERT architecture, capturing biophysical properties of protein sequences through self-supervised learning

ProstT5 is a protein language model capable of translating between protein sequences and structures.

A protein sequence pre-training model based on T5-3B architecture that captures biophysical properties through self-supervised learning

ESM-2 is a state-of-the-art protein model trained on masked language modeling objectives, suitable for fine-tuning on various protein sequence input tasks.

A BERT-based protein sequence pre-training model that extracts biophysical features through self-supervised learning from 2.1 billion protein sequences

ESM-1b is a Transformer-based protein language model trained via unsupervised learning on protein sequence data, capable of predicting protein structure and function.

RiNALMo is a non-coding RNA (ncRNA) model pre-trained based on the masked language modeling (MLM) objective, trained through self-supervised learning on a large number of non-coding RNA sequences.

A BERT-based model for predicting RNA torsion angles and pseudo-torsion angles from RNA sequences

ESM-2 is a state-of-the-art protein model trained with masked language modeling objectives, suitable for fine-tuning on various protein sequence tasks.

ProtGPT2 is a protein language model based on the GPT2 architecture, capable of generating novel protein sequences while retaining key features of natural proteins.

Helix-mRNA is a hybrid model combining state-space and transformer architectures, specifically designed for mRNA sequence analysis

ESM++ is a faithful implementation of ESMC, supporting batch operations and fully compatible with the Huggingface ecosystem without relying on the official ESM Python package. The large version corresponds to ESMC's 600-million parameter model.

DRP is a model focused on robot motion planning, trained for 7k epochs (approximately 7M steps) under a 5M checkpoint

RNA foundation model pre-trained on non-coding RNA data using masked language modeling (MLM) objective

ESM++ is a faithful implementation of ESMC, supporting batch processing and compatible with the standard Huggingface interface without requiring the ESM Python package. The small version corresponds to the 300-million-parameter version of ESMC.

SaProt is a pre-trained model based on protein sequence and structural information, specifically optimized for low pLDDT regions.

This model generates embedding vectors for protein sequences, supporting shortened embeddings to accelerate search tasks.

Distilled version of ProtBert-UniRef100 model for protein feature extraction and downstream task fine-tuning

This model is fine-tuned based on Bert-Base-Uncased and is used for the multi-label classification task of predicting protein functions based on protein amino acid sequences.

GENERator is a generative genome foundation model with a 98,000 base pair context length and 3 billion parameters, trained on an extended dataset of eukaryotic DNA

The ultimate tool for codon optimization, capable of converting protein sequences into DNA sequences optimized for target organisms.

This model is used to predict mass spectrometry data for given amino acid sequences.

ProLLaMA is a protein large language model based on the Llama-2-7b architecture, specializing in multitask protein language processing.

ProtT5-XL-BFD is a self-supervised pre-trained model based on protein sequences, using the T5 architecture, trained on 2.1 billion protein sequences for protein feature extraction and downstream task fine-tuning.

GoBERT is a model specifically designed for general gene function prediction. It can effectively capture the relationships between Gene Ontology (GO) functions by leveraging gene ontology graph information.

GPN-MSA is a language model trained on human and 89 other vertebrate species, specializing in DNA sequence analysis and variant effect prediction.

The first de novo generator of linear peptide binders dependent solely on target protein sequences

ProLLaMA is a protein large language model based on Llama-2-7b, specializing in multitask protein language processing, capable of performing protein sequence generation and superfamily classification tasks.

Optimized model based on HIV-BERT for predicting the collection site of HIV V3 loop samples, with significantly improved accuracy compared to the original model

A protein function prediction tool based on the Evolutionary Scale Modeling (ESM), which predicts the molecular function of proteins through amino acid sequences.

A protein molecular function prediction model based on the Gene Ontology (GO) and the ESM2 architecture

The ProtST framework enhances the pre-training and understanding of protein sequences through biomedical text, constructs the ProtDescribe dataset, designs three pre-training tasks, and supports supervised learning and zero-shot prediction.

Isoformer is a model capable of accurately predicting differential transcript expression, outperforming existing methods and fully leveraging multimodal data.

A distilled version of ProtGPT2, compressed into a more efficient small model through knowledge distillation, maintaining performance while improving inference speed

AbLang is an antibody language model specifically designed for processing antibody sequences, particularly heavy chain sequences. The model is trained on uppercase amino acid letters and is suitable for protein feature extraction and downstream task fine-tuning.

Ankh3 is a protein language model based on the T5 architecture. It is pre - trained by jointly optimizing masked language modeling and sequence completion tasks, and is suitable for protein feature extraction and sequence analysis.

nanoBERT is a nanobody-specific transformer for predicting amino acids at specific positions in a given query sequence.

SegmentNT-multi-species is a segmentation model based on the Nucleotide Transformer, designed to predict the locations of various genomic elements at single-nucleotide resolution.