Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Binary file added source/_static/structure-prediction/protenix.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
15 changes: 15 additions & 0 deletions source/web-app/structure-prediction/using-structure-prediction.rst
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@ We recommend using:
- ESMFold for predictions that must be completed quickly.
- MiniFold is a fast single-sequence structure prediction model built on ESM-2, delivering accuracy comparable to ESMFold while reducing inference time by 10–20×. It is designed for rapid prediction of large numbers of protein structures and currently supports single-chain proteins.
- RosettaFold-3 is a three-track neural network for protein structure and complex prediction, useful for modeling protein-protein interactions and supporting experimental structure determination.
- Protenix predict the 3D structure of biological molecules including proteins, DNA, RNA, and small molecule ligands, as well as how they interact with each other.

Accessing the Structure Prediction tool
---------------------------------------
Expand Down Expand Up @@ -112,6 +113,20 @@ The **Advanced Options** section contains several parameters:
.. image:: /_static/structure-prediction/rosettafold.png
:alt: RosettaFold-3

Using Protenix
-----------------
When using Protenix, you can enter or upload multiple proteins in the input fields provided.

The **Advanced Options** section contains several parameters:

- **Number of recycles** Controls how many times the model feeds its predicted structure back into itself for refinement. Higher values improve accuracy but increase computation time.
- **Diffusion Samples** Sets how many independent structure candidates are generated per input. More samples increase the chance of finding the best conformation but proportionally increase runtime.
- **Sampling Steps** Defines the number of denoising steps the diffusion process takes to build each structure. More steps produce more precise, physically valid outputs at the cost of longer inference.
- **Step Scale** Adjusts the sampling temperature, controlling how broadly the model explores structural space. Higher values produce more diverse candidates; lower values give tighter, more consistent results.

.. image:: /_static/structure-prediction/protenix.png
:alt: Protenix

Visualizing your sequence
--------------------------

Expand Down
Loading