ProteInfer

ProteInfer predicts functional properties of protein amino acid sequences using deep convolutional neural networks (CNNs) to assign Enzyme Commission (EC) numbers and Gene Ontology (GO) terms.

Key Features:

• Model architecture: Uses deep convolutional neural networks (CNNs) to infer protein function from sequence data.
• Predicted outputs: Produces Enzyme Commission (EC) numbers and Gene Ontology (GO) terms as functional annotations.
• Alignment-free inference: Infers function directly from amino acid sequences without relying on sequence alignment or database comparisons.
• Sequence representation: Maps full-length amino acid sequences into a generalized functional space for downstream analysis.
• Computational efficiency: Implements a computationally efficient approach for function prediction.
• Local execution: Supports performing all computations locally on a user's personal computer, avoiding data upload to remote servers.

Scientific Applications:

• Protein function annotation: Assigns EC numbers and GO terms to protein sequences for functional characterization.
• Complementary annotation: Provides alignment-free predictions that can complement traditional alignment-based annotation methods.
• Downstream analysis: Enables embedding of full-length sequences into a functional space to support downstream interpretation and analysis.

Methodology:

Applies deep convolutional neural networks (CNNs) to full-length amino acid sequences, mapping sequences into a generalized functional space to predict EC numbers and GO terms without sequence alignment or database comparisons, with computations executable locally.