GREMA

GREMA (Gene networks Reconstruction using Evolutionary Modelling Algorithm) is a program for inferring a novel type of gene regulatory network (GRN) with confidence levels for every inferred regulation, which is emulated GRN (eGRN). The higher the confidence level, the more accurate the inferred regulation. GREMA gradually determines the regulations of an eGRN with confidence levels in descending order using either an S-system or a Hill function-based ordinary differential equation model. It makes use of an evolutionary modelling algorithm (EMA) that is based on evolutionary intelligence, including both crowd wisdom and an evolutionary strategy, to cope with the underdetermined problem. EMA uses an intelligent genetic algorithm to solve the large-scale parameter optimisation problem.

Citing GREMA

GREMA: Modelling of emulated gene regulatory networks with confidence levels based on evolutionary intelligence to cope with the underdetermined problem. Bioinformatics, 2020, PMID:32399550

Input Data Format

There are 2 input tsv files: 1) Time-series data and 2) Domain knowledge data.

The format of time-series data (e.g., Dream4 insilico_size10_1)

• repeat_number= (e.g.,repeat_number=5)
• timepoint_number= (e.g.,timepoint_number=21)
• time-series profile (repeat_number <tab> gene_name <tab> gene_expression)

The format of domain knowledge data (e.g., Dream 4 insilico size10_1)

• TF
• GENE
• REGULATORY (+:activation, -:repression, ?:unknown, 0:No regulation).

Getting start

git clone https://github.com/NctuICLab/GREMA.git
cd GREMA
cd EMA_HFODE;make

Usage of GREMA

To check the usage of GREMA_main.pl

perl GREMA_main.pl -h
Usage: GREMA_main.pl [Options]
Options:
       -i      [FILE]  time-series profile
       -o      [PATH]  Output Directory
       -k      [FILE]  knowledge of regulatory
       -m      [model] {s-system,HFODE} Default is HFODE
       -f      [Fitness] {0,1,2,3,4} Default is 1
       -c      [cc]    {0,1} Default is 0, consider correlation coefficient
       -g      [Generation] Number of generation, default is 1000
       -t      [No]    Number of threads
       -min    [value] ex: 0.01 [min value] default is 0
       -h      Show the usage

Run GREMA

perl GREMA_main.pl \
    -i input/Dream4_10_1_timeseries_expression.txt \
    -k input/insilico_size10_no_prior_knowledge.txt \
    -o output/Dream4_10_1/ \
    -t 10

Results of GREMA

less -S output/Dream4_10_1/final_results.txt

The format of results of GREMA (e.g., Results of Dream4 size10 Network1: final results)

• TF
• GENE
• REGULATORY (+:activation, -:repression, ?:unknown, 0:No regulation).
• CONFIDENCE_LEVEL

Evaluation of GREMA

Installing Scikit learn for evaluation script

Mac OSX

pip3 install -U numpy scipy scikit-learn

Linux

sudo apt-get install build-essential python3-dev python3-setuptools \
                     python3-numpy python3-scipy \
                     libatlas-dev libatlas3gf-base
sudo pip3 install -U scikit-learn

To check the usage of evaluation

python3 evaluation/dream4_evaluate.py --help
usage: Evaluation of GREMA performance. [-h] [-d DIRECTED] [-u UNDIRECTED] input
positional arguments:
  input                 Prediction file (final_results.txt) with path.

optional arguments:
  -h, --help            show this help message and exit
  -d DIRECTED, --directed DIRECTED
                        The gold-standard directed GRN file with path
  -u UNDIRECTED, --undirected UNDIRECTED
                        The gold-standard undirected GRN file with path

GREMA can reconstruct the directed GRNs, so we can evaluate the results using directed or undirected gold standard GRNs. Here are two types of evaluating commands:

Run the evaluation script for undirected gold standard GRN.

python3 evaluation/dream4_evaluate.py \
    -u evalutation/gold_standards_undirected/10/DREAM4_GoldStandardUndirected_InSilico_Size10_1.tsv \
    output/Dream4_10_1/final_results.txt
==============================
Start evaluating the results of GREMA
Undirected golden file is: evalutation/gold_standards_undirected/10/DREAM4_GoldStandardUndirected_InSilico_Size10_1.tsv
Prediction file is output/Dream4_10_1/final_results.txt
loading gold file:evalutation/gold_standards_undirected/10/DREAM4_GoldStandardUndirected_InSilico_Size10_1.tsv
loading predict results:output/Dream4_10_1/final_results.txt
==============================
Performance of GREMA:
TP=11, TN=23, FP=9, FN=2
Sensitivity=0.846
Specificity=0.719
Accuracy=0.756
ROC AUC=0.782
PR AUC=0.720
==============================

Run the evaluation script for directed gold standard GRN.

python3 evaluation/dream4_evaluate.py \
  -d evalutation/gold_standards/10/DREAM4_GoldStandard_InSilico_Size10_1.tsv \
  output/Dream4_10_1/final_results.txt
==============================
Start evaluating the results of GREMA
Directed golden file is: evalutation/gold_standards/10/DREAM4_GoldStandard_InSilico_Size10_1.tsv
Prediction file is output/Dream4_10_1/final_results.txt
loading gold file:evalutation/gold_standards/10/DREAM4_GoldStandard_InSilico_Size10_1.tsv
loading predict results:output/Dream4_10_1/final_results.txt
==============================
Performance of GREMA:
TP=11, TN=57, FP=18, FN=4
Sensitivity=0.733
Specificity=0.760
Accuracy=0.756
ROC AUC=0.747
PR AUC=0.579
==============================

Contact

• Shinn-Ying Ho: syho@nctu.edu.tw
• Ming-Ju Tsai: milutsai.bi98g@g2.nctu.edu.tw
• Jyun-Rong Wang: bengdex.bi99g@g2.nctu.edu.tw