Drop prediction files or folders here (.zip, .gz, .xz, .cif, .pdb, .json, .npz, .npy, .csv) or click to browse
p53–MDM2 examples:
3- to 4-chain complexes (AlphaFold3):
6- to 14-chain complexes (AlphaFold3):
AlphaFold3 predicted via AlphaFold Server. ColabFold, Boltz-2, Chai-1, OpenFold3, and Protenix-v2 predicted via Tamarind Bio. ESMFold2 predicted via biohub API. ORC–Cdc6, CCT/TRiC, eIF3, and Complex I are higher-order complexes from the FlyPredictome interactome network, shipped as lightweight bundles.
For batch analysis, use lis.py (PPI scoring) and lis_to_cxc.py (ChimeraX script generation) from AFM-LIS.
Rows are sorted by iLIS. A model's rank in its name (rank_001 …) is the predictor's own order (ipTM-based for AlphaFold and ColabFold), so it can differ from the iLIS order. Click a row to switch models; this updates all visualizations and the script below.
Thresholds based on large-scale Y2H reference sets in yeast, fly, and human predicted using ColabFold. See Kim et al., 2026, for iLIS benchmark details. Different platforms may require different thresholds.
Chain identity
Looks up each chain in UniProt by sequence and lists any domains found.
contact Local Interaction Score (cLIS) Maps (PAE ≤ 12 Å & Cβ ≤ 8 Å)
0contact Local Interaction Score (cLIS)1
Predicted Aligned Error (PAE)
Static PAE plot(s) supplied with this precomputed bundle (numeric PAE was dropped to reduce size; scores were precomputed by lis.py).
Sequence Viewer
Light background = LIR (PAE ≤ 12 Å). Dark background with white text = cLIR (PAE ≤ 12 Å & Cβ ≤ 8 Å). Each chain uses its own color.
Linear Contact Map
Lines connect cLIR residue pairs within Cβ distance in the 3D structure. Color gradient from chain A to chain B.domains:
Circular Contact Map
domains:label direction
Arcs represent chains. Lines connect cLIR residue pairs in physical contact across all chain pairs.
Selected Model
All Models Combined
3D Structure Viewer
Interactive 3D preview powered by Mol* (Sehnal et al. 2021). LIR regions shown as cartoons; cLIR residues highlighted. Drag to rotate, scroll to zoom. If the viewer appears empty, refresh the page.
Chains shown in 3D & chord:
Visualization Script
Fill gaps ≤residues|Min segment ≥residues
Gap filling bridges short breaks for continuous cartoon. Min segment removes isolated LIR fragments shorter than the threshold.
Both scripts show every subunit, each chain's LIR/cLIR colored to match the 3D viewer (LIR light & cLIR dark in Gradient shade; one color per chain in Solid). .cxc · full structure opens Full structure = the whole complex. .cxc · partial structure opens Partial structure = the same subunit composition, each chain trimmed to the LIR segments shown above (Fill gaps / Min segment honored). Distinct filenames, so nothing overwrites; residue numbers preserved.
Download a script and its matching structure file into the same folder, then open the script in ChimeraX or PyMOL.
Per-Residue Intra-chain LIS & pLDDT
Per-residue metrics for each chain across all models. Selected model as solid line; other models as light traces; average as dashed line.
predicted DockQ from the mean pLDDT of the interface and the number of contacting residue pairs (Cβ ≤ 8 Å), with no PAE (Bryant et al., 2022)
pDockQ2
pDockQ with each contact weighted by its PAE, per chain (Zhu et al., 2023)
LIpDockQ
the pDockQ formula computed only on cLIR pairs (PAE ≤ 12 Å and Cβ ≤ 8 Å), so contacts the PAE does not support no longer count (Kim & Perrimon, 2026)
LIpDockQ2
the pDockQ2 formula on cLIR pairs only, so a small confident interface is not diluted by a large, poorly placed contact surface (Kim & Perrimon, 2026)
Because their constants were fitted to DockQ on the original interfaces, LIpDockQ and LIpDockQ2 rank predictions; they are not DockQ estimates.
PAE (Predicted Aligned Error) measures how confidently the model predicts the relative position of two residues — lower values indicate higher confidence.