HMI-PRED 2.0 (Host-Microbe Interaction PREDictor version 2.0) is a web service
for 3D structure-based protein-protein interaction prediction between human and any microbial species.
The rationale behind HMI-PRED2 is the
interface mimicry, which is shown in the animated figure above.
A microbial protein may have a surface patch similar to that of another protein (mimicked) that is known to interact with a human (host) protein.
By mimicking the surface (interaction interface), the microbial protein can interact with the host protein, even when the global structures differ signifcantly.
HMI-PRED 2.0 performs several steps to predict host-microbe interactions via
1. Surface matching. The surface patches of input microbe protein are structurally compared to the interaction interfaces in the known protein-protein interaction database using TM-align.
2. Interface filtering. The known interactors (potentially targetted by the microbial protein) of successfully aligned (mimicked) chains are collected, considering structural clashes with the microbe protein. At this stage, interface hot spots are matched between microbe and mimic chains, where higher matches suggest interface mimicry is evolutionarily better conserved for the microbe.
3. Docking. To ensure the host-microbe interactions are energetically favorable, Rosetta docking is performed for all potential host-microbe protein pairs that passed the filtering stage.
4. Refinement. Docked HMIs are refined to yield final predicted complexes. The final predicted HMIs are thus geometrically well-aligned in terms of interface mimicry, conserved evolutionary information, and energetically favorable.
External analyses. The predicted host-microbe interactions form a protein-protein interaction network with structural details. Network analysis can be performed using String, an external tool from EMBL
HMI-PRED 2.0 comes with several improvements on top of its predecessor, HMI-PRED (version 1).
1. More interfaces. HMI-PRED 2.0 comes with a larger protein-protein interaction interface data set to cover more structural space. HMI-PRED 2.0 interface data set (template database) covers PDB structure deposited by January 2021 and keep adding more. Since both versions of HMI-PRED rely on the interface data to predict HMIs, it is important to have more up-to-date interface data set. More interface templates means HMI-PRED 2.0 can predict more interactions.
2. Supports mmCIF-structure files. Old .pdb format files cannot represent large protein structures with more than 100K atoms, while such proteins are more common now. HMI-PRED 2.0 supports mmCIF-formatted structure files to break the atom count limit.
3. Network analysis module. Many projects involve multiple protein-protein interactions for microbe proteins. HMi-PRED 2.0 provides a way to connect the predicted interactions to human protein-protein interactions via network analysis module (external analysis).
4. Improved visualization. Major changes have been made to the structure visualization parts as well as the appearance of the website. HMI-PRED 2.0 provides visually appealing and controllable 3D structure views using NGL-viewer. Interactive statistics plots are generated using Plotly.js.
5. Advanced search functions. Under HMI-PRED 2.0 Library, users can now make advanced queries to search for predicted HMIs. Templates can also be searched under the library. More information and examples are on the tutorial page.