isSPA_1.1.2

In this version, several scripts are added to do the preprocess and postprocess.

Update

1. We previously observed that isSPA might produce inconsistent results on different machines with varying GPU cards and CUDA versions. This issue has now been partially resolved.

Experimental tests show that this version identifies more correct particles and achieves higher resolution under the same data processing methods (including Class3D, Refine3D, Postprocess, and other subsequent processing).

2. In this version, when norm_type=1 is in the config, the computation is performed on the entire image. Experimental tests indicate that the results with norm_type=1 are better than those with norm_type=0, with more particles identified. When norm_type=0 is used, the large image is divided into smaller (e.g.720x720) sections for computation, and the normalization algorithm is not applied. This setting is suitable for machines with less powerful GPUs, as processing large images can be demanding on the GPU, and omitting normalization can speed up the computation.

Installation

1. Download HDF5 package from the official website: https://support.hdfgroup.org/downloads/index.html
2. Uncompress it, such as

tar -xzf hdf5-1.14.5.tar.gz

4. Install HDF5 according to ./hdf5-1.14.5/release_docs/INSTALL_Autotools.txt.
5. (Recommended) Add the absolute path of the library of HDF5 to LD_LIBRARY_PATH. For instance,

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/user/Software/hdf5/lib

6. Enter the directory of isSPA_1.1.2, and modify LIB_HDF5 and INCLUDE_HDF5 in the Makefile according to the installation paths in step 3.
7. Edit the first line of Makefile, making sure the path of SHELL is correct in your system.
8. Execute the following commands (N is the number of available threads):

make -j N
make install

9. (Recommended) Add the absolute paths of ./isSPA_1.1.2/build and ./isSPA_1.1.2/isSPA_scripts to environment variables. For example,

export PATH=/home/user/Software/isSPA_1.1.2/build:$PATH
export PATH=/home/user/Software/isSPA_1.1.2/isSPA_scripts:$PATH

Answers to some frequently asked questions

1. Which version should be used to generate projection files? You can use EMAN1 or EMAN2 now.
2. What version of Python should I use? Python 3.
3. How does isSPA use multiple GPUs? You can split the entire dataset into multiple parts using the first and last parameters in the config file. Each part can then be processed on a separate GPU.

Contributor

Li Rui, Chen Yuanbo, Zhao Mingjie, Cheng Yuanhao