biocentral

Biocentral provides a free and open embedding-based service that enables researchers to generate protein language model embeddings, obtain embedding-based protein feature predictions, and train embedding-based models.

Key Features:

• Embedding Generation: Supports standardized access to most protein language models currently in use for generating protein embeddings.
• Protein Feature Prediction: Provides embedding-based protein feature predictions through a predict module.
• Model Training: Enables training of embedding-based models through a dedicated training module.
• Programmatic Access: Offers an Application Programming Interface (API) for programmatic access to services.

Scientific Applications:

• Large-Scale Protein Analysis: Supports large-scale analysis of protein databases, as demonstrated with the BFVD virus database.
• Method Reproduction: Facilitates reproduction of existing embedding-based prediction methods.
• Protein Prediction Research: Enables experimental and computational biologists to leverage protein language model embeddings without requiring expensive hardware or expert knowledge.

Methodology:

Biocentral utilizes protein language models to generate embeddings, which serve as protein representations for downstream feature prediction and model training.