$\mathrm{C_T G}$ program

$\mathrm{C_T G}$ (Hi-C To Geometry) algorithm is a diffusion-based method to obtain reliable geometric information of the chromatin from Hi-C data. Here we provide a program based on CUDA, which can rapidly calculate the $\mathrm{C_T G}$ distance from the Hi-C data.

Installation and usage

This program is based on CUDA. Before installation, please make sure that you have installed CUDA 10.1 or later versions.
Download the file CTG.cu.
Makefile. Run this command in the terminal:
```
 nvcc CTG.cu -o CTG -lcublas
```
Add the path with CTG into your environment variables.
Use this command to run the program:
```
 CTG -i input_matrix_name -o output_matrix_name -alpha alpha -iteration_numbers k
```
with the meanings of the parameters:
1. -i: The name of the file storing the input matrix. It should be stored in the coordinate format:
```
 bin1    bin2    count

 e.g. 

 2	2	4.78862e+01
 2	3	4.79380e+01
 2	5	3.41081e+01
 2	8	1.97129e+01
 2	9	9.91527e+00
 2	10	2.68997e+00
 2	11	8.69303e+00
 2	12	1.12322e+01
 2	13	5.88250e+00
 2	14	1.05338e+01
```
  Note that the index of bins should start from 1.
2. -o: The name of the file to store the output $\mathrm{C_T G}$ matrix. It should be a binary file, and can be read using the following code in python:
```
 import numpy as np
 CTG = np.fromfile(output_matrix_name,dtype='float32')
 N = int(len(CTG)**(1/2))
 CTG = np.reshape(CTG,(N,N))
```
3. -alpha: The parameter in the $\mathrm{C_T G}$ algorithm. It is $\lambda$ in the following equation:
  
  $$S^{(k)} = \sum_{t=1}^k \mathrm{e}^{-\lambda t} P^t$$
  
  The default value is 0.3.
4. -iteration_numbers: The largest k in the equation above. The default value is 20.

PKUGaoGroup/CTG

$\mathrm{C_T G}$ program

Installation and usage