# Bootstrapping

Use this logic to resolve structural heterogeneity on the 3D level, by reconstructing multiple 3D volumes from randomly combined 2D images or class averages.

## Usage

Data sets of particle images often contain different structural states of the same complex. Bootstrapping allows to detect such structural differences by first creating random subgroups of images and then computing structures for each of these subgroups.

- The randomly drawn images are grouped in sets. Both the number of sets and the number of images per set can be specified by the user. Depending on the data one will have to find an optimum for the number of particles used for each subgroup: Smaller numbers will facilitate identification of lowly-populated structural states, but also result in very low signal-to-noise ratios of the 3D structures.
- For each set one can then compute a 3D structure (enable option:
**UseReconstruction**). In this case the images need assigned angular orientations. In case of class sums, either i) angular reconstitution can be used to assign angular orientations (enable option:**Use Euler**) or ii) one can run the euler logic with all images as input to get a consensus model and to preassign euler angels for the images based on this consensus model. In case of single particle images, one may assign angles based on alignment by projection matching against a consensus structure. - It is mandatory to have assigned euler angles if
**Use Euler**is disabled. - The resulting 3D volumes typically have a low signal-to-noise ratio. Therefore, one will generally have to perform subsequently a Classification of the 3D structures to average similar structures in order to increase the signal-to-noise ratio.

## Process

Parameters | Description |
---|---|

Use 'euler' - No | No angular assignment will be performed; 3D reconstruction will require pre-assigned angular information from header values. |

Use 'euler' - Use Euler | Angular assignment will be performed by angular reconstitution. |

→ Symmetry | Symmetry to be used for angular assignment. |

Use reconstruction - Use reconstruction | 3D structures will be computed; requires either prior angular assignment or enabling the “Use Euler” option. |

→ Symmetry | Symmetry to be used for 3D reconstruction. |

Use reconstruction - No | No 3D structures will be computed; images will only by resampled into subsets. |

Project to one half-sphere | Useful to improve angular sampling: When checked images with angles on the southern hemisphere of the Euler sphere will be mirrored to their equivalent on the northern hemisphere. |

Number of sets | Number of sets of images to be randomly drawn. |

Size of single set | Number of images per subset |

Input | Description |
---|---|

Input images | Stack of 2D images, class averages to be combined into random sets |

Output | Description |
---|---|

Random sets | Randomly drawn subsets of images. |

3Ds | 3D structures computed from the individual subsets. Only available if UseReconstruction is enabled. |

New/Changed Header Values | Description |
---|---|

randomSet | Number of the random set the image belongs to |