SecBic-CCA: Secured Biclusterings - Cheng and Church Algorithm: privacy-preserving gene expression data analysis by biclustering algorithm -- Cheng and Church algorithm -- over gene expression data (i.e., yeast cell cycle) with Homomorphic Encryption operations such as sum, or matrix multiplication in Python under the MIT license.
We apply Pyfhel as a python wrapper for the Microsoft SEAL library and biclustlib, a library for biclustering algorithms.
First you need to ensure that all packages have been installed.
- See
requirements.txt
- numpy>=1.23.1
- setuptools>=65.5.0
- pandas>=1.5.0
- scikit-learn>=1.1.1
- Pyfhel>=3.3.1
- matplotlib>=3.5.2
- scipy>=1.9.0
- munkres>=1.1.4
You can clone this repository:
> git clone https://github.com/ShokofehVS/SecBic-CCA.git
If you miss something you can simply type:
> pip install -r requirements.txt
If you have all dependencies installed:
> pip3 install .
To install Pyfhel, on Linux,gcc6
for Python (3.5+
) should be installed. (more information regarding installation of Pyfhel )
> apt install gcc
Biclustering or simultaneous clustering of both genes and conditions as a new paradigm was introduced by Cheng and Church's Algorithm (CCA). The concept of bicluster refers to a subset of genes and a subset of conditions with a high similarity score, which measures the coherence of the genes and conditions in the bicluster. It also returns the list of biclusters for the given data set.
Our input data is yeast Saccharomyces cerevisiae cell cycle taken from Tavazoie et al. (1999) which was used in the orginal study by Cheng and Church;
To measure the similarity of encrypted biclusters with non-encrypted version, we use Clustering Error (CE) as an external evaluation measure that was proposed by Patrikainen and Meila (2006);
SecBiclib/
software library contains algorithms, datasets, evaluation and sample scriptsalgorithms/
original Cheng and Church algorithm and its encrypted counterpartdatasets/
real (yeast cell cycle) and synthetic data generationevaluation/
external evaluation measures including CEscripts
small programs to showcase functionalities of SecBic library
To run the sample implementation of Cheng and Church algorithm:
> python3 cheng_church_yeast.py
import time
from SecBiclib.algorithms import ChengChurchAlgorithm
from SecBiclib.datasets import load_yeast_tavazoie
import numpy as np
m0 = time.perf_counter()
# load yeast data used in the original Cheng and Church's paper
data = load_yeast_tavazoie().values
# missing value imputation suggested by Cheng and Church
missing = np.where(data < 0.0)
data[missing] = np.random.randint(low=0, high=800, size=len(missing[0]))
# creating an instance of the ChengChurchAlgorithm class and running with the parameters
cca = ChengChurchAlgorithm(num_biclusters=5, msr_threshold=996.0, multiple_node_deletion_threshold=1.2)
biclustering = cca.run(data)
print(biclustering)
m1 = time.perf_counter()
print("Time Performance in Original Algorithm: ", round(m1 - m0, 5), "Seconds")
To run the sample implementation of Secured Cheng and Church algorithm:
> python3 secured_cheng_church_yeast.py
import time
from SecBiclib.algorithms import SecuredChengChurchAlgorithm
from SecBiclib.datasets import load_yeast_tavazoie
import numpy as np
m0 = time.perf_counter()
# load yeast data used in the original Cheng and Church's paper
data = load_yeast_tavazoie().values
# missing value imputation suggested by Cheng and Church
missing = np.where(data < 0.0)
data[missing] = np.random.randint(low=0, high=800, size=len(missing[0]))
# creating an instance of the SecuredChengChurchAlgorithm class and running with the parameters
secca = SecuredChengChurchAlgorithm(num_biclusters=5, msr_threshold=996.0, multiple_node_deletion_threshold=1.2)
biclustering = secca.run(data)
print(biclustering)
m1 = time.perf_counter()
print("Time Performance in Calculating Homomorphically: ", round(m1 - m0, 5), "Seconds")
To run the sample implementation of external evaluation measure (i.e., CE) on original and encrypted one:
> python3 evaluation.py
from SecBiclib.algorithms import ChengChurchAlgorithm
from SecBiclib.algorithms import SecuredChengChurchAlgorithm
from SecBiclib.evaluation import clustering_error
from SecBiclib.datasets import load_yeast_tavazoie
# load yeast data used in the original Cheng and Church's paper
data = load_yeast_tavazoie().values
num_rows, num_cols = data.shape
# creating an instance of the ChengChurchAlgorithm, SecuredChengChurchAlgorithm classes
# and running with the parameters of the original study
cca = ChengChurchAlgorithm(num_biclusters=5, msr_threshold=996.0, multiple_node_deletion_threshold=1.2)
secca = SecuredChengChurchAlgorithm(num_biclusters=5, msr_threshold=966.0, multiple_node_deletion_threshold=1.2)
biclustering_ref = cca.run(data)
biclustering_pre = secca.run(data)
# creating an instance of the clustering error class
# and running with reference and predicted algorithms
ce_eval = round(clustering_error(biclustering_pre, biclustering_ref, num_rows, num_cols),5)
print(ce_eval)