Need the input? Generate it with lis.py from AFM-LIS.
📄
lis.py output *+ FASTA
Drop lis.py output and FASTA here — CSV, XLSX, FASTA, or zips. Sorted automatically.
🧬
FASTA optional · or drop above
Predicted sequences → color cLIR on the 3D structure and fill the sequence viewer (you can also drop these in the panel on the left)
🏷️
Identity map optional
name, chain + ID columns (uniprot / gene / label — any scheme) · resolves arbitrary prediction names
🧊
Structure optional
Drag or click to upload a structure (PDB/CIF) to use instead of AlphaFold DB
Proteins not resolving? Drop the FASTA used for the prediction — proteins are matched by exact sequence, so a truncated or engineered chain will not resolve from its name alone. Otherwise fill in the identity-map template below, or use auto-convert. Tested with ColabFold inputs. Still stuck? Open an issue on GitHub.
3D structure via AlphaFold DB — resolved from FASTA sequence (exact) or protein name + length.
Resolved from the names in your file, so it might land on the wrong protein or species. Not right?
0partners0predictions0clusters0query length
Cluster colors— applies to all panels (heatmap, frequency, 3D, scatter)
Contact residue frequency
SVG w · h · font · x-ticks
How often each query residue is a contact residue (cLIR) across predictions (each model of each partner; one per partner when that box is ticked), colored by the dominant cluster · click clusters to filter. Cn (N): N = predictions (AlphaFold ranks) in that cluster. Tracks below the bars: AlphaFold pLDDT (0–100) — ≤50 50–70 70–90 >90 AM path. = AlphaMissense mean pathogenicity per residue (averaged over all ~19 substitutions at that position; blue = benign → red = pathogenic; human only, from AlphaFold DB).
Clustered interaction fingerprint
SVG w · h · font
Rows = partners (dendrogram order), columns = query residues. Dark blue = cLIR contact, very light = none; left: dendrogram + cluster strip.
Cluster info
Partners in each cluster (dendrogram order) — (proteins / predictions). A partner can have several AlphaFold ranks (each prediction's own model order, ipTM-based, not the iLIS order), so it may appear at multiple rows and across clusters. With one prediction per partner ticked, each partner is clustered once, on its best-iLIS prediction.
Interaction Residues
The interaction residues on each protein, taken from the lis.py indices. No structure file needed. LIR = local interaction residue (PAE ≤ 12 Å); cLIR = contact residue (PAE ≤ 12 Å & Cβ ≤ 8 Å).
Query protein structure, residues colored by the cluster that consensus-contacts them · click clusters to isolate. Cn (N): N = predictions (AlphaFold ranks) in that cluster.
Highlight residues contacted by ≥ % of a cluster's members (consensus / core residues).
Color only clusters with ≥ predictions (hides tiny clusters).
(the (N) on each cluster chip — AlphaFold ranks, not unique partners).
Color selection:clusterpLDDTAlphaMissense· AM pathogenicity average ≥
Interaction scatter plot
gray = below cutoff / unclustered · colored = clustered. rank = global rank (1..N) by the Y-axis metric across all predictions.
Partners
* actifpTM is approximate: lis.py's estimate from the PAE over residue pairs in contact (About).