InterPLM
This dashboard helps explore features extracted from the protein language model ESM-2-8M via Sparse Autoencoders (SAEs). Details on training, analysis, and visualizations in the InterPLM paper. Select ESM embedding layer Select random feature Or specify SAE feature number View Protein Activations Customize the protein selection and visualization below. Defaults to max activated proteins (a): Select activation group Number of proteins to visualize Enter Uniprot IDs Enter protein sequences Highlight high activation residues (Highlighting selected feature f/8386) The feature activates on bacterial hexapeptide repeat transferase domains involved in lipid A and peptidoglycan biosynthesis, particularly detecting conserved acetyl-CoA binding sites. No Swiss-Prot concepts found for this feature.
This dashboard helps explore features extracted from the protein language model ESM-2-8M via Sparse Autoencoders (SAEs). Details on training, analysis, and visualizations in the InterPLM paper. Select ESM embedding layer Select random feature Or specify SAE feature number View Protein Activations Customize the protein selection and visualization below. Defaults to max activated proteins (a): Select activation group Number of proteins to visualize Enter Uniprot IDs Enter protein sequences Highlight high activation residues (Highlighting selected feature f/8386) The feature activates on bacteria
Explore this link on the map →