MOLECULAR SPARSE REPRESENTATION BY 3D ELLIPSOID RADIAL BASIS FUNCTION NEURAL NETWORKS

Volumetric electron density maps

Volumetric electron density maps are often modelled as the volumetric Gaussian density maps $\phi : \mathbb{R}^3 \rightarrow \mathbb{R}$. The definition of the volumetric Gaussian density maps is as follows,

$$\phi(x) = \sum_{i} exp(-d(||x-xi||^2 - ri^2)$$

where the parameter $d$ is positive and controls the decay rate of the kernel functions, $x_i$ and $r_i$ are the location and radius of atom $i$.

---

Instructions for MOLECULAR SPARSE REPRESENTATION (under Linux):

1. Unpack SparseGaussianMolecule.tar.gz The SparseGaussianMolecule contains these folders:

   - "\pqr": directory for input .pqr files.
   - "\output": directory for output .txt files.
   - "\example": directory for a script main.py.
   

   (In main.py: The variable PQR is the file name of input .pqr file. It should be in the \pqr folder. The variable d is the speed rate in Gaussian surface.)

2. Execute main.py.

3. After the main.py is ended, the output .txt file is under the "\output" directory.

Notice:

You need to install

- Python version: python36
- Minimun required version of CUDA: 9.2
- Minimun required verison of PyTorch: 1.3
- numpy and torchvision

before run the code "main.py".

---

A quick guide for usage of this version:

command line style:

• cd example

• python3 main.py xxx.pqr d

(Notice: the pqr file should be put in the folder named "\pqr". The output .txt file is in the folder named "\output". )

Example: python3 main.py adp.pqr 0.5

---

You may contact Dr. Sheng Gui, email: shenggui@lsec.cc.ac.cn or Ms. Zhaodi Chen, email: 20174207022@stu.suda.edu.cn, if you have any problem running this program.