As illustrated in Ripstein et al. (eLife 2017;10.7554/eLife.25754) ab-initio or reference-free 3D classification of particle images can yield information on previously unknown conformations and intermediates which can shed light on protein mechanisms.
Ripstein et al. ran 93,469 particle images through cryoSPARC's ab-initio reconstruction algorithm multiple times using between two and six ab-initio classes to identify a second conformation of VAT that was significantly different from the previously known six-fold symmetric stacked-ring conformation.
This second conformation, representing approximately 15% of the particle images, showed the VAT complex in the process of unfolding a neighbouring complex as substrate in a 'hand-over-hand' mechanism.
Gold-standard refinement of the six-fold symmetric conformation (applying C6 symmetry) in cryoSPARC yielded a resolution of 3.9 Å, and refinement of the substrate-engaged class (no symmetry) yielded a resolution of 4.8 Å.