Protein intelligence

ProtLoc-AI

Prediction Mutation analysis Explainability
◎

Localization prediction

Residue-level localization across 11 compartments with ESM-2 embeddings.

◇

Variant effect analysis

Quantify how mutations shift predicted localization and clinical risk.

◆

Residue interpretability

Attention maps and signal motifs for transparent model behavior.

Architecture

⌁Protein sequence

Raw amino acids as input

◎ESM-2 language model

650M parameters · 250M sequences

◇Residue embeddings

1280 dimensions per position

◆Attention classifier

Learned pooling over residues

◎Three outputs
Prediction
Variant analysis
Interpretability
0
Proteins
0
Compartments
0
AUROC
0
Parameters

Built with ESM-2 · PyTorch · FastAPI

Predict localization

11 compartments · residue attention · ESM-2 650M

Length: 0
Attention analysis▼

Variant effect analysis

Predict how mutations alter protein localization

Length: 0

Format: [original][position][mutant], comma separated

Signal analysis▼
Attention comparison▼

Protein relocalizationExperimental

AI-guided sequence design

This feature is in active development. Results are experimental and may vary.
Length: 0