IPSAE Now Available on Vecura

What is IPSAE?

IPSAE (Interprotein Scoring for AlphaFold/Boltz Evaluations) is a scoring function designed to evaluate the reliability of interprotein contacts predicted by AlphaFold2, AlphaFold3, and Boltz1/2. It takes as input the Predicted Aligned Error (PAE) matrix and the 3D structure file of a multimeric complex, then computes a series of metrics that assess the quality of the interfaces between chains. These metrics include ipSAE, ipTM, pDockQ, pDockQ2, and LIS, which provide chain-pair and residue-level scores. It helps users to quickly and accurately assess the confidence in predicted inter-chain interactions, without requiring any model weights or GPU resources. It is especially useful for researchers who need to post-process large sets of multimer predictions and require detailed information on the quality of chain-chain interfaces.

What can users do with IPSAE on Vecura?

With IPSAE on Vecura, users can:

• Evaluate the reliability of chain-chain interfaces in their multimer predictions.
• Obtain per-chain-pair and per-residue scores for each interface in the predicted complex.
• Visualize the confidence of each residue in the structure using PyMOL coloring scripts.
• Analyze the predicted 3D structures from AlphaFold2, AlphaFold3, and Boltz1/2 without setting up additional software or infrastructure.
• Use the provided scores to guide further experimental validation or computational analysis.

What the output means

The output provides a set of scores and metrics that indicate the quality and confidence of the interfaces between chains in the multimeric complex. This includes chain-pair level scores such as ipSAE, ipTM, pDockQ, pDockQ2, and LIS, as well as a per-residue table with pSAE, pTM_pae, and d0 values. The PyMOL script allows for visual inspection of the confidence at each residue, aiding in the interpretation of the results. This output should be used to support scientific decision making. It does not replace experimental validation.

Why this matters

The ability to reliably predict and assess the quality of protein-protein interfaces is crucial for understanding the molecular basis of many biological processes and for the design of new therapeutics. By providing a rapid and accurate method for evaluating these interfaces, IPSAE enables researchers to make more informed decisions about which predictions to pursue experimentally, thereby accelerating the pace of discovery.

• Developed by: Dunbrack Lab at Fox Chase Cancer Center
• Source: Official GitHub repo (DunbrackLab/IPSAE)
• Reference: IPSAE preprint (bioRxiv 2025)