Available tools

Structure prediction

The majority of users take advantage of our structure prediction tools because our jobs can run on large-memory GPU cards on NSF supercomputers. This means that COSMIC2 can predict structural assemblies with up to 5,500 amino acids, whereas Google Colab notebooks are much smaller in size).

• AlphaFold2 Full stack for the ground-breaking structure prediction software

• ColabFold Faster implementation of AlphaFold2 by using MMSeqs for sequence search.

• ESMFold Language model-based structure prediction. Fast, no sequence alignment needed.

• IgFold A structure prediction tool designed specifically for antibody heavy/light chain complexes.

• OmegaFold Language model-based structure prediction. Like ESMFold, no sequence alignment required. ​

Model building

Users can utilize Model Angelo for building atomic models into density maps with resolutions of 3.5 Å or lower. They can also use Model Angelo in a predictive mode where it can output hidden Markov models of its 'best guess' for a amino acid sequences.

• Model Angelo Automatically builds structural models into density maps. Will only work for data that is better than 3.5Å resolution.

• HHBlits Will take outputs from Model Angelo to provide a hidden Markov model of possible sequences that match a given 3D reconstruction.

Single particle cryo-EM

We have tools for single particle analysis (e.g., RELION) in addition to 3D reconstruction modification (e.g., DeepEMhancer, AR-DECON). We also have the ability to run 3D conformational variability analysis software (e.g., cryoDRGN).

• ​Single particle analysis 

  • RELION Single particle analysis for 2D & 3D classification, 3D refinement, multibody refinement, and post-processing.

  • cryoDRGN Deep-learning based approach to analyze and reconstruct conformationally variable cryo-EM datasets.

  • csparc2star.py Conversion of cryoSPARC output files into RELION formatting.

• Map modification & analysis

  • AR-DECON  Deconvolution to restore cryo-EM maps with anisotropic resolution.

  • DeepEMhancer Generative deep-learning tool to improve the interpretability of density maps.

  • Efficiency  Software that estimates how much of Fourier space is missing from a given Euler angle file and suggests what tilt angle is needed for future data collection.  

  • Local B-factor / LocSpiral / LocOccupancyThis will either enhance a given map or provide estimates of local B-factors given a 3D reconstruction or provide occupancy estimates.

Cryo-electron tomography

We are beginning to offer cryo-electron tomography software. At the moment, we have IsoNet, which offers a deep-learning based generative approach to mitigate the effects of the missing wedge on cryo-ET reconstructions. 

• IsoNet A generative deep learning program that modifies tomographic reconstructions to dampen noise and mitigate effects of the missing wedge.