References

Version and Activity

Analysis

Social

How to use

You can use the sys package to import the pygenmet package.

import sys

sys.path.insert(', 'pygement_path')
import pygenmet

Introduction

Nonalcoholic fatty liver disease (NAFLD) is a consequence of defects in diverse metabolic pathways or nutritional effects that involve a hepatic accumulation of triglycerides. Features of these deviations might determine whether NAFLD progresses to nonalcoholic steatohepatitis (NASH). We investigated whether the diverse defects observed in patients with NAFLD are due to different NAFLD subtypes with specific serum metabolomic profiles and whether these can distinguish patients with NASH from patients with simple steatosis.

We collected liver and serum from several murine models, which have genetic alterations, nutritional alterations or both. We also analyzed serum metabolomes of 535 patients with biopsy-proven NAFLD (353 with simple steatosis and 182 with NASH) and compared them with serum metabolomes of these murine models. In a previous analysis, we identified several subtypes of NAFLD using cluster analysis and we recognized markers that differentiate steatosis from NASH in each subtype.

In this study, a specific Genetic Algorithm (GA) has been developed to improve the identification of these subtypes in a human cohort and create a scalable methodology to identify subtypes of NAFLD. The final aim of this work is to develop a methodology for precision care in NAFLD.

Genetic Algorithms are a very useful tool to solve optimization problems but they need a good definition of a fitness function. This package has been developed in Python language and NumPy Style Python Docstrings has been used to document the package.

example numpy style python docstrings
numpydoc docstring guide

FIGURE: Genetic algorithm scheme and functions associated with each step.

INITIAL POPULATION/POPULATION
- _get_father(): See tools:_get_father function.
- _get_population(): See tools:_get_population function.
FITNESS
- _get_fitness(): See fitness:_get_fitness function.
- _get_fitness_robust(): See fitness:_get_fitness_robust function.
- _get_fitness_l(): See fitness:_get_fitness_l function.
SELECTION
- _selection(): See operators:_selection function.
CROSSOVER
- _crossover(): See operators:_crossover function.
- _arithmetic_crossover(): See operators:_arithmetic_crossover function.
MUTATION
- _mutation(): See operators:_mutation function.

FIGURE: Genetic Algorithms Applied to Translational Strategy in NASH. Learning from Mouse Models. Poster presented in the II Jornadas Doctorales de la UPV/EHU 2019.

FIGURE: Genetic Algorithms Reveal Potential to Stratify MASLD Heterogeneity. Poster presented in the Liver Meeting, Boston Nov. 10-14, 2023.

CLASS `chromosome`

One of the contributions of this package is the chromosome class as it is specifically designed to easily solve this optimization problem.

Import the pygenmet package

from pygenmet import *

With Pyhton a classcan be created. A class provides a lot of functionalities. We can create a new clas and then we can create a new type of object in Python. This new class allows new instances and these instances can also have methods.

The class chromosome has been created to optimizate the algorithm. The docstring associated to the chromosome class is the following.

?chromosome

The class chromosome has a copy of its genes, its fitness and 
the number of No NAFLD samples and NAFLD samples for print purposes.

Attributes
----------

Methods
-------

print()  
  Prints the chromosome

to_json()   
    
Notes
-----
The representation of a chromosome is the following way
    
0110101101 ... 0110101101 || 0110110011 ... 0110110011     6     6     0.89456        
-------------------------    -------------------------   ----- -----   -------
            A                            B                 C     D        E
                
Description of the representation:
  A] Wild Type Genes
  B] Knock-Out Genes
  C] Number of Wild Type cases selected 
  D] Number of Knock-out cases selected
  E] The fitness value. The value of the adaptation for this chromosome
    
Examples
--------
>>> CH = chromosome('01101011010110110011', 0.8945, 10, 10)
>>> print("Genes: {}".format(CH.Genes))
Genes: 01101011010110110011        
>>> print("WT Genes: {}".format(CH.WT_Genes))
WT Genes: 0110101101        
>>> print("KO Genes: {}".format(CH.KO_Genes))
KO Genes: 0110110011        
>>> print("Fitness: {}".format(CH.Fitness))
Fitness: 0.8945        
>>> CH.print()

CH = chromosome('0110101101011000111101101011010110001111', 0.4125, 10, 30)

print("Genes: {}".format(CH.Genes))

Genes: 0110101101011000111101101011010110001111

print("WT Genes: {}".format(CH.WT_Genes))

WT Genes: 0110101101

print("KO Genes: {}".format(CH.KO_Genes))

KO Genes: 011000111101101011010110001111

print("Fitness: {}".format(CH.Fitness))

Fitness: 0.4125

Methods

A Python method is like a Python function, except it is associated with object/classes. Methods in python are very similar to functions except for two major differences:

The method is implicitly used for an object for which it is called.
The method is accessible to data that is contained within the class.

The chromosome class has currently two methods:

print()
to_json()

print()

The print() method allows to show the chromosome information on the screen or in a file.

The print() method has an argument, N, and with this argument we can customized the number of alleles to print. This method print first de Wild Types genes and second the KO genes separated by ||. The number of alleles equal to 1 for each gen is also printed. At last, the fitness function value is shown.

CH.print(N = 3)

011 ... 101 || 011 ... 111 	   6	  18	         0.4125

CH.print(N = 5)

01101 ... 01101 || 01100 ... 01111 	   6	  18	         0.4125

to_json()

json or JavaScript Object Notation is an open standard file format, that uses human-readable text to store and transmit data objects consisting of attribute–value pairs and array data types. It is a very common data format, with a diverse range of applications.

CH.to_json(path_or_buf = 'out.json')

file out.json has de following structure:

{"chromosome": [{"Genes": "0110101101011000111101101011010110001111", 
                 "Fitness": 0.4125, 
                 "WT_Genes": "0110101101", 
                 "KO_Genes": "011000111101101011010110001111"}]}

fitness

Import the pygenmet package

from pygenmet import *

Currently, three fully customizable fitness functions are developed in pygenmet package:

_get_fitness
_get_fitness_robust
_get_fitness_l

fitness._get_fitness

The docstring associated to the _get_fitness function is the following.

?_get_fitness

Parameters
----------
genes : Genes of a chromosome
nwt : Number of observations of Wild Type
nko : Number of observations of Knock-out
dwt : Data of observations of Wild Type
dko : Data of observations of Knock-out
obj : Array of fold-changes of metabolites in murine model
l : float Usually 0.5 / length(fc1). If length(fc1) = 50 then l = 0.01

Returns
-------
p : float
    The aptitude of a genes of a chromosome (solution) is the correlation between fc1 and fc2. 
    Initially the aptitude (fitness) of a chromosome is the Pearson Correlation between the 
    fold-change of Knock-out select samples and Wild Type selected samples with the fold-change of
    the murine model analyzed (a vector of ¿50? fold-changes).

Notes
-----
This function evaluates the Pearson Correlation 
between fc1 and fc2 with a penalty for different sign 
in the values. This function uses pearsonr from scipy and
sign from numpy to evaluate the aptitude.

The penalty has been named l.

Examples
--------

fitness._get_fitness_robust

The docstring associated to the _get_fitness_robust function is the following.

?_get_fitness_robust

Parameters
----------
genes : Genes of a chromosome
nwt : Number of observations of Wild Type
nko : Number of observations of Knock-out
dwt : Data of observations of Wild Type
dko : Data of observations of Knock-out
obj : Array of fold-changes of metabolites in murine model
l : float Usually 0.5 / length(fc1). If length(fc1) = 50 then l = 0.01

Returns
-------
p : float
    The aptitude of a genes of a chromosome (solution) is the correlation between fc1 and fc2. 
    Initially the aptitude (fitness) of a chromosome is the Pearson Correlation between the 
    fold-change of Knock-out select samples and Wild Type selected samples with the fold-change of
    the murine model analyzed (a vector of ¿50? fold-changes).

Notes
-----
This function evaluates the Pearson Correlation 
between fc1 and fc2 with a penalty for different sign 
in the values. This function uses pearsonr from scipy and
sign from numpy to evaluate the aptitude.

The penalty has been named l.

Examples
--------

fitness._get_fitness_l

The docstring associated to the _get_fitness_l function is the following.

?_get_fitness_l

Parameters
----------
genes : Genes of a chromosome
nwt : Number of observations of Wild Type
nko : Number of observations of Knock-out
dwt : Data of observations of Wild Type
dko : Data of observations of Knock-out
obj : Array of fold-changes of metabolites in murine model

Returns
-------
p : float
    The aptitude of a genes of a chromosome (solution) is the correlation between fc1 and fc2. 
    Initially the aptitude (fitness) of a chromosome is the Pearson Correlation between the 
    fold-change of Knock-out select samples and Wild Type selected samples with the fold-change of
    the murine model analyzed (a vector of ¿50? fold-changes).

Notes
-----
This function evaluates the Pearson Correlation 
between fc1 and fc2 with a penalty for different sign 
in the values. This function uses pearsonr from scipy and
sign from numpy to evaluate the aptitude.

The penalty has been named l.

Examples
--------

operators

Import the pygenmet package

from ..tools.tools import chromosome

import pandas as pd
import numpy as np

Crossover

operators._crossover

The docstring associated to the _crossover function is the following.

?_crossover

Parameters
----------
chromosome_a: Chromosome
chromosome_b: Chromosome
k: k-point. If k = 1 the method will be single-point crossover, 
            If k = 2 the method will be two-point crossover. 
            If k = N (with N the length of chromosome) the method will be the uniform crossover. Its value is 1 by default.

Returns
-------
Two new chromosomes

Notes
-----
[1]

References
----------
[1] https://en.wikipedia.org/wiki/Crossover_(genetic_algorithm)

Examples
--------
>>> A = _get_father(nwt = dwt.shape[0], nko = dko.shape[0], 
                    dwt = dwt, dko = dko,
                    geneSet = '01', 
                    get_fitness = _get_fitness, obj = np.array(FC.loc[:, murine_models_in[4]]))

>>> B = _get_father(nwt = dwt.shape[0], nko = dko.shape[0], 
                    dwt = dwt, dko = dko,
                    geneSet = '01', 
                    get_fitness = _get_fitness, obj = np.array(FC.loc[:, murine_models_in[4]]))

>>> A.print()
011000111111010 ... 001000010001110 || 001010011000110 ... 011001001111111 	  43	 336	0.03363686825576087

>>> B.print()
100101111101111 ... 010001000100101 || 010101001001010 ... 111010000101010 	  42	 345	0.1410663212679354

>>> A1, B1 = _crossover(A, B, get_fitness = _get_fitness,  
                    nwt = dwt.shape[0], nko = dko.shape[0], dwt = dwt, dko = dko, 
                    obj = np.array(FC.loc[:, murine_models_in[4]]), k = 1)

>>> A1.print()
011000111111010 ... 001000010001110 || 001010011000110 ... 111010000101010 	  43	 327	0.007288743389698732

>>> B1.print()
100101111101111 ... 010001000100101 || 010101001001010 ... 011001001111111 	  42	 354	0.15370276754212897

>>> A1, B1 = _crossover(A, B, get_fitness = _get_fitness,  
                    nwt = dwt.shape[0], nko = dko.shape[0], dwt = dwt, dko = dko, 
                    obj = np.array(FC.loc[:, murine_models_in[4]]), k = 772)

>>> A1.print()
110000111101111 ... 000000010001111 || 011111001001110 ... 011000001111111 	  43	 355	0.017131591198246282

>>> B1.print()
001101111111010 ... 011001000100100 || 000000011000010 ... 111011000101010 	  42	 326	0.13633106854522628

operators._arithmetic_crossover

?_arithmetic_crossover

Notes
-----
For this specific genotype (GeneSet = '01') the arithmetic crossover techniques are available.
These techniques are:
  [1] AND
  [2] OR
  [3] XOR
Arithmetic crossover - some arithmetic operation is performed to make a new offspring [1]

References
----------
[1] https://www.obitko.com/tutorials/genetic-algorithms/crossover-mutation.php

Mutation

operators._mutation

?_mutation

Parameters
----------
father : Genes of the father
geneSet : Set of genes availables. Usually '0' and '1'
get_fitness : Function to improve
nwt : Number of No NAFLD Samples
nko : Number of NAFLD Samples
dwt : Data of No NAFLD Samples
dko : Data of NAFLD Samples
obj : The fold-changes of murine model to adjust
type : Type of mutation process, 'bsm' by default. Five types are allowed.
       bf : bitflip
       bsm : bitstringmutation
       sw : swap
       in : inversion
       sc : scramble
p : Parameter to modify the posibility of change in a genes

Returns
-------
A new mutated chromosome by the type of mutation process

Notes
-----
The flipbit (fb) mutation is an aggressive mutation because change all genes. 
If in a specific gen the alelo value is '0' this process will change it to '1'.

fb: '0101010101' -> '1010101010'

The bitstringmutation (bsm) mutation process will change a gen with a 
probability 'p'. If p = None then p = 1/len(genes) [1].

bsm: '0101010101' -> '0101110101'

References
----------

[1] https://en.wikipedia.org/wiki/Mutation_(genetic_algorithm)
[2] https://www.tutorialspoint.com/genetic_algorithms/genetic_algorithms_mutation.htm

Examples
--------

Selection

operators._replacement

?_replacement

Parameters
----------
type : Type of replacement
       rws : Replace Worst Strategy
       rts : Restricted Tournament Selection
       wams : Worst Among Most Similar Replacement


Notes
-----    
The _replacement function is ised to change the old generation population with the new generation. 
Four process have been evaluated [1]:

 1) Replace Worst Strategy (RW). The worst element of the population is replaced if the child improves it.
 It offers high selective pressure, even when its parents are chosen randomly.

 2) Restricted Tournament Selection (RTS) [2].

 3) Replace the Worst Among Most Similar Replacement (WAMS) [3]. the worst chromosome of the
 set of the N (N = 3,...) parents most similar to the offspring.

 4) Deterministic Crowding Algorithm (Deterministic Crowding, DC) [4]. To facilitate the comparison we will use in our experiments. A variant of the DC in which a single descendant will be generated for each cross, which will replace the most similar parent if it improves it. 

References
----------
[1] https://sci2s.ugr.es/keel/pdf/keel/congreso/4-diversidadfinal2_daniel_molina.pdf
[2] G.   Harik.   Finding   multimodal   solutions   using restricted  tournament  selection. Proc.  6th  Int.  Conf. Genetic Algorithms, páginas 24-31, 1995.
[3] W. Cedeño and V. Vemuri. Multi-niche crowding in genetic algorithms and its application to the assembly of dna restriction-fragments. Evolutionary Computation, 2(4):321-345, 1995.    
[4] S.W. Mahfoud. Crowding and preselection revised. Parallel Problem Solving from Nature 2, páginas 27-36, 1992.

operators._selection

?_selection

THERE ARE TWO TYPES OF SELECTION. THE ONE THAT SELECTS FUTURE PARENTS AND THE ONE THAT SELECTS PARENTS FOR CROSSING THEM

Matching Techniques:
Parents can be selected in a way that maintains population diversity.

1) Prohibition of Crossbreeding Based on Ancestry. An individual cannot mate with themselves, their parents, their children, or their siblings.

2) Incest Prohibition. Two parents mate if their Hamming distance is above a certain threshold.

3) Diverse Pairing. An individual mates with another that is quite different. Hamming distance.

THE ONE THAT SELECTS A NEW GENERATION

1) Random Selection (RS)

2) Tournament Selection (TS): It chooses the individual with the best fitness from
among N individuals selected randomly (N = 2, 3, . . . ) Tournament selection is a widely used
parent selection procedure in which the idea is to randomly choose a number of individuals from
the population, tournament size, (with or without replacement), select the best individual from
this group, and repeat the process until the number of selected individuals matches the population size.
Usually, the tournament size is 2, and in such a case, a probabilistic version has been used
in which the selection of individuals is allowed without them necessarily being the best.

3) Linear Rank Selection (LRS): The population is ranked based on its fitness, and a selection probability
is associated with each individual that depends on their order.

4) Roulette Selection (RS): Assigns a selection probability proportional to the individual's fitness value.
Baker (1987) introduces a method called stochastic universal sampling, which uses a single spin of the
roulette wheel with circular sectors proportional to the objective function.
Individuals are selected from markers, equally spaced and with random starting points.

5) Elitist: In the elitist selection model, the best individual in the population at
time t is forced to be selected as a parent.

Parameters
----------
chromosomes : List of chromosomes of the population
size : Number of individuals to select from the population. This parameter is not considered with method = 'elitist'.
N : Number of Genes to print (if trace is True)
trace : Should be the chromosomes and the selection printed on the screen
method: The method used for selection: ['random', 'tournament', 'linear', 'roulette', 'elitist']

Returns
-------
fitness : The fitness of each chromosome (individual)
psel : Probabilities of selecction of each chromosome. pesl(x) = f(x) / sum(f(y)), for each x in y
selecction : List of selected chromosomes

Notes
-----
The aim of this function is to evaluate the fitness of each population (by the chromosome structure) and 
to asign a probability to select for a new population (next generation).

Examples
--------
>>> A_ = []
>>> A_.append(chromosome(genes = '01010001010110100100', nwt = 10, nko = 10, fitness = 2.3454))
>>> A_.append(chromosome(genes = '01011011110110100100', nwt = 10, nko = 10, fitness = 1.4401))
>>> A_.append(chromosome(genes = '01010001010100000100', nwt = 10, nko = 10, fitness = 0.9254))
>>> A_.append(chromosome(genes = '01110101010110100100', nwt = 10, nko = 10, fitness = 7.1104))
>>> A_.append(chromosome(genes = '01010101110111100100', nwt = 10, nko = 10, fitness = 2.1494))
>>> A_.append(chromosome(genes = '01010001110000100100', nwt = 10, nko = 10, fitness = 1.9954))
>>> A_.append(chromosome(genes = '00000000010110100100', nwt = 10, nko = 10, fitness = 1.7323))
>>> A_.append(chromosome(genes = '01010000111110100111', nwt = 10, nko = 10, fitness = 1.5002))
>>> A_.append(chromosome(genes = '01110011010110100111', nwt = 10, nko = 10, fitness = 2.4119))
>>> A_.append(chromosome(genes = '01010011011110101100', nwt = 10, nko = 10, fitness = 2.1414))

>>> selection, fit, ps, sel_mean, sel_maximun = _selection(A_, size = 4, N = 8, trace = True, method = 'roulette')

|---------------------------------------------------------------------------------------------|
|-------------- fenotypes ---------------||--- genotypes ---||--- fitness ---||- Probability -|
01010001...01000101 || 01101001...10100100         4        4           2.3454       0.09912204
01011011...01101111 || 01101001...10100100         7        4           1.4401       0.07244166
01010001...01000101 || 01000001...00000100         4        2           0.9254       0.05727278
01110101...11010101 || 01101001...10100100         6        4           7.1104       0.23955286
01010101...01010111 || 01111001...11100100         6        5           2.1494       0.09334567
01010001...01000111 || 00001001...00100100         5        2           1.9954       0.08880708
00000000...00000001 || 01101001...10100100         1        4           1.7323       0.08105318
01010000...01000011 || 11101001...10100111         4        7           1.5002       0.07421289
01110011...11001101 || 01101001...10100111         6        6           2.4119       0.10108189
01010011...01001101 || 11101011...10101100         5        6           2.1414       0.09310990
|---------------------------------------------------------------------------------------------|
|-------------- fenotypes ---------------||--- genotypes ---||--- fitness ---||- Probability -|
01110101...11010101 || 01101001...10100100         6        4           7.1104       0.23955286
01010001...01000101 || 01101001...10100100         4        4           2.3454       0.09912204
01010001...01000111 || 00001001...00100100         5        2           1.9954       0.08880708
01010011...01001101 || 11101011...10101100         5        6           2.1414       0.09310990

>>> selection, fit, ps, sel_mean, sel_maximun = _selection(A_, size = 4, N = 8, trace = True, method = 'linear')

|---------------------------------------------------------------------------------------------|
|-------------- fenotypes ---------------||--- genotypes ---||--- fitness ---||- Probability -|
01010001...01000101 || 01101001...10100100         4        4           2.3454       0.15555555
01011011...01101111 || 01101001...10100100         7        4           1.4401       0.02222222
01010001...01000101 || 01000001...00000100         4        2           0.9254              0.0
01110101...11010101 || 01101001...10100100         6        4           7.1104              0.2
01010101...01010111 || 01111001...11100100         6        5           2.1494       0.13333333
01010001...01000111 || 00001001...00100100         5        2           1.9954       0.08888888
00000000...00000001 || 01101001...10100100         1        4           1.7323       0.06666666
01010000...01000011 || 11101001...10100111         4        7           1.5002       0.04444444
01110011...11001101 || 01101001...10100111         6        6           2.4119       0.17777777
01010011...01001101 || 11101011...10101100         5        6           2.1414       0.11111111
|---------------------------------------------------------------------------------------------|
|-------------- fenotypes ---------------||--- genotypes ---||--- fitness ---||- Probability -|
01011011...01101111 || 01101001...10100100         7        4           1.4401       0.02222222
01010000...01000011 || 11101001...10100111         4        7           1.5002       0.04444444
01110101...11010101 || 01101001...10100100         6        4           7.1104              0.2
01010001...01000101 || 01101001...10100100         4        4           2.3454       0.15555555

>>> selection, fit, ps, sel_mean, sel_maximun = _selection(A_, size = 4, N = 8, trace = True, method = 'random')

|---------------------------------------------------------------------------------------------|
|-------------- fenotypes ---------------||--- genotypes ---||--- fitness ---||- Probability -|
01010001...01000101 || 01101001...10100100         4        4           2.3454              0.1
01011011...01101111 || 01101001...10100100         7        4           1.4401              0.1
01010001...01000101 || 01000001...00000100         4        2           0.9254              0.1
01110101...11010101 || 01101001...10100100         6        4           7.1104              0.1
01010101...01010111 || 01111001...11100100         6        5           2.1494              0.1
01010001...01000111 || 00001001...00100100         5        2           1.9954              0.1
00000000...00000001 || 01101001...10100100         1        4           1.7323              0.1
01010000...01000011 || 11101001...10100111         4        7           1.5002              0.1
01110011...11001101 || 01101001...10100111         6        6           2.4119              0.1
01010011...01001101 || 11101011...10101100         5        6           2.1414              0.1
|---------------------------------------------------------------------------------------------|
|-------------- fenotypes ---------------||--- genotypes ---||--- fitness ---||- Probability -|
01010101...01010111 || 01111001...11100100         6        5           2.1494              0.1
01010001...01000101 || 01000001...00000100         4        2           0.9254              0.1
01010011...01001101 || 11101011...10101100         5        6           2.1414              0.1
01010001...01000101 || 01101001...10100100         4        4           2.3454              0.1

>>> selection, fit, ps, sel_mean, sel_maximun = _selection(A_, size = 4, N = 8, trace = True, method = 'elitist')

|---------------------------------------------------------------------------------------------|
|-------------- fenotypes ---------------||--- genotypes ---||--- fitness ---||- Probability -|
01010001...01000101 || 01101001...10100100         4        4           2.3454       ----------
01011011...01101111 || 01101001...10100100         7        4           1.4401       ----------
01010001...01000101 || 01000001...00000100         4        2           0.9254       ----------
01110101...11010101 || 01101001...10100100         6        4           7.1104       ----------
01010101...01010111 || 01111001...11100100         6        5           2.1494       ----------
01010001...01000111 || 00001001...00100100         5        2           1.9954       ----------
00000000...00000001 || 01101001...10100100         1        4           1.7323       ----------
01010000...01000011 || 11101001...10100111         4        7           1.5002       ----------
01110011...11001101 || 01101001...10100111         6        6           2.4119       ----------
01010011...01001101 || 11101011...10101100         5        6           2.1414       ----------
|---------------------------------------------------------------------------------------------|
|-------------- fenotypes ---------------||--- genotypes ---||--- fitness ---||- Probability -|
01010101...01010111 || 01111001...11100100         6        5           2.1494       ----------
01010001...01000101 || 01101001...10100100         4        4           2.3454       ----------
01110011...11001101 || 01101001...10100111         6        6           2.4119       ----------
01110101...11010101 || 01101001...10100100         6        4           7.1104       ----------

>>> selection, fit, ps, sel_mean, sel_maximun = _selection(A_, size = 4, N = 8, trace = True, method = 'tournament')

|---------------------------------------------------------------------------------------------|
|-------------- fenotypes ---------------||--- genotypes ---||--- fitness ---||- Probability -|
01010001...01000101 || 01101001...10100100         4        4           2.3454       ----------
01011011...01101111 || 01101001...10100100         7        4           1.4401       ----------
01010001...01000101 || 01000001...00000100         4        2           0.9254       ----------
01110101...11010101 || 01101001...10100100         6        4           7.1104       ----------
01010101...01010111 || 01111001...11100100         6        5           2.1494       ----------
01010001...01000111 || 00001001...00100100         5        2           1.9954       ----------
00000000...00000001 || 01101001...10100100         1        4           1.7323       ----------
01010000...01000011 || 11101001...10100111         4        7           1.5002       ----------
01110011...11001101 || 01101001...10100111         6        6           2.4119       ----------
01010011...01001101 || 11101011...10101100         5        6           2.1414       ----------
|---------------------------------------------------------------------------------------------|
|-------------- fenotypes ---------------||--- genotypes ---||--- fitness ---||- Probability -|
00000000...00000001 || 01101001...10100100         1        4           1.7323       ----------
01010000...01000011 || 11101001...10100111         4        7           1.5002       ----------
01110011...11001101 || 01101001...10100111         6        6           2.4119       ----------
01010101...01010111 || 01111001...11100100         6        5           2.1494       ----------
01110101...11010101 || 01101001...10100100         6        4           7.1104       ----------

tools

Import the pygenmet package

from pygenmet import *

import pandas as pd
import numpy as np

tools._get_father

With the function \_get\_father() we can to generate an initial solution. We can to initializate nwt, dwt, nko and dko. For replicate purpose, this function has the random_state argument.

The docstring associated to the _get_father function is the following.

?_get_father

Parameters
----------
nwt : Number of observations of Wild Type
nko : Number of observations of Knock-out
dwt : Data of observations of Wild Type
dko : Data of observations of Knock-out
geneSet: Usually a list as '01'
get_fitness: The fitness function to use
obj : Array of fold-changes of metabolites in murine model
random_state: If you don't mention the random_state in the code, then whenever you execute your code a new random value

Returns
-------
This function returns a chromosome

Examples
--------    
>>> _get_father(nwt = nwt, nko = nko, dwt = dwt, dko = dko, geneSet = '01', get_fitness = _get_fitness, obj = FC.FC).print()
011101101010000 ... 000000001000001 || 101110011001010 ... 011110110000011    37     261    -0.04568152517704738

>>> _get_father(nwt = nwt, nko = nko, dwt = dwt, dko = dko, geneSet = '01', get_fitness = _get_fitness, obj = FC.FC).print()
000101011001011 ... 101100010101101 || 010000101000100 ... 110111010001010    46     252    -0.07192848665245967

>>> _get_father(nwt = nwt, nko = nko, dwt = dwt, dko = dko, geneSet = '01', get_fitness = _get_fitness, obj = FC.FC).print()
110101100111011 ... 000110110110111 || 110010000110100 ... 110010011101011    49     267    -0.11718322152992591

nwt = 15
nko = 25
dwt = pd.DataFrame(np.random.rand(nwt,10))
dko = pd.DataFrame(np.random.rand(nko,10))

for i in np.arange(10):
    A = _get_father(nwt = nwt, nko = nko, dwt = dwt, dko = dko, 
                    geneSet = '01', get_fitness = _get_fitness, 
                    obj = [1,0,1,-1,1,0,0,1,0.5, 0.5])

    A.print()

000100010101011 ... 000100010101011 || 010011110000011 ... 000110111000101 	   6	  12	-0.01750455939922404
010100000111001 ... 010100000111001 || 000111111010001 ... 100010010110110 	   6	  13	0.01583165234542458
001010100111101 ... 001010100111101 || 101110011011111 ... 111110100001101 	   8	  15	0.1337101625593648
011011101001110 ... 011011101001110 || 011111111111101 ... 111010000110111 	   9	  18	0.3326538149497444
101100111001001 ... 101100111001001 || 010100110110000 ... 100001010110001 	   8	  11	0.09708129202192302
010001010010001 ... 010001010010001 || 100111000100111 ... 001110100100100 	   5	  11	0.1442876222067664
000000000011101 ... 000000000011101 || 100100011111100 ... 111000011110101 	   4	  14	0.28826914052937097
110100111110100 ... 110100111110100 || 111011001010111 ... 101111010101001 	   9	  15	-0.377007167796123
100111000110010 ... 100111000110010 || 111001110110000 ... 100001101111100 	   7	  15	0.3216335311818925
000111101101000 ... 000111101101000 || 100010011101100 ... 011001100001110 	   7	  12	-0.09813075927314555

for i in np.arange(10):
    A = _get_father(nwt = nwt, nko = nko, dwt = dwt, dko = dko, 
                    geneSet = '01', get_fitness = _get_fitness, 
                    obj = [1,0,1,-1,1,0,0,1,0.5, 0.5], 
                    random_state = 42)

    A.print()

100011100001001 ... 100011100001001 || 101101100100011 ... 000111111011111 	   6	  17	0.04061614908467767
100011100001001 ... 100011100001001 || 101101100100011 ... 000111111011111 	   6	  17	0.04061614908467767
100011100001001 ... 100011100001001 || 101101100100011 ... 000111111011111 	   6	  17	0.04061614908467767
100011100001001 ... 100011100001001 || 101101100100011 ... 000111111011111 	   6	  17	0.04061614908467767
100011100001001 ... 100011100001001 || 101101100100011 ... 000111111011111 	   6	  17	0.04061614908467767
100011100001001 ... 100011100001001 || 101101100100011 ... 000111111011111 	   6	  17	0.04061614908467767
100011100001001 ... 100011100001001 || 101101100100011 ... 000111111011111 	   6	  17	0.04061614908467767
100011100001001 ... 100011100001001 || 101101100100011 ... 000111111011111 	   6	  17	0.04061614908467767
100011100001001 ... 100011100001001 || 101101100100011 ... 000111111011111 	   6	  17	0.04061614908467767
100011100001001 ... 100011100001001 || 101101100100011 ... 000111111011111 	   6	  17	0.04061614908467767

tools._get_population

The docstring associated to the _get_population function is the following.

?_get_population

Parameters
----------
individuals : Number of individual of first generation or initial population
nwt : Number of observations of Wild Type
nko : Number of observations of Knock-out
dwt : Data of observations of Wild Type
dko : Data of observations of Knock-out
geneSet : Usually a list as '01'
get_fitness : The fitness function to use
obj : Array of fold-changes of metabolites in murine model
random_state : If you don't mention the random_state in the code, then whenever you execute your code a new random value

Returns
-------
population_ : A list of the initial population (chromosomes) to initialize the algorithm

Notes
-----

Examples
--------
>>> nwt = 15
>>> nko = 25
>>> dwt = pd.DataFrame(np.random.rand(nwt,10))
>>> dko = pd.DataFrame(np.random.rand(nko,10))
>>> A = _get_population(individuals = 10, 
                        nwt = nwt, nko = nko, dwt = dwt, dko = dko, 
                        geneSet = '01', get_fitness = _get_fitness, obj = [1,0,1,-1,1,0,0,1,0.5, 0.5])

>>> [A_.print() for A_ in A];
110101101011111 ... 110101101011111 || 101110000011110 ... 111101111110101    11      16    0.27010320047631353
110001101000111 ... 110001101000111 || 111011001000100 ... 001001100010011     8      12    0.4153593539526977
110110110000111 ... 110110110000111 || 111001000001001 ... 010011110110101     9      13    0.24120026823063395
001001101100010 ... 001001101100010 || 101000001100111 ... 001110111010010     6      12    0.5976992970173265
110110101000111 ... 110110101000111 || 100011000000110 ... 001100011111011     9      12    0.4327563879291974
010110111011011 ... 010110111011011 || 101101110101111 ... 011110010010000    10      13    0.19109063637942147
011011111001110 ... 011011111001110 || 000001110011000 ... 110001111010111    10      13    0.19291545113399683
100010011100001 ... 100010011100001 || 011101010010100 ... 101001111100011     6      14    0.3838922708272582
110001111111010 ... 110001111111010 || 110111010001010 ... 010100000000100    10       9    0.04833875255418793
001111011101100 ... 001111011101100 || 100011000100101 ... 001010101000110     9      10    0.6180057110908423

nwt = 15
nko = 25
dwt = pd.DataFrame(np.random.rand(nwt,10))
dko = pd.DataFrame(np.random.rand(nko,10))
A = _get_population(individuals = 10, 
                    nwt = nwt, nko = nko, dwt = dwt, dko = dko, 
                    geneSet = '01', get_fitness = _get_fitness, 
                    obj = [1,0,1,-1,1,0,0,1,0.5, 0.5])

[A_.print() for A_ in A];

001100101100011 ... 001100101100011 || 010111011100111 ... 001111010101001 	   7	  15	0.22717953288324988
011111001110011 ... 011111001110011 || 110011011010001 ... 100011100110100 	  10	  13	-0.14158703160537597
000111000110010 ... 000111000110010 || 000111011011101 ... 111010101101111 	   6	  16	-0.14752436179984593
111110011001001 ... 111110011001001 || 111101011001011 ... 010111111010101 	   9	  17	0.37230832910925965
111100011100110 ... 111100011100110 || 001000100111010 ... 110100000100011 	   9	   9	0.09942842284405343
011100001011101 ... 011100001011101 || 001010010010100 ... 101000010010000 	   8	   7	-0.29440398210415947
111011011011101 ... 111011011011101 || 000101001001101 ... 011011010111110 	  11	  13	0.262667504755609
000000010010100 ... 000000010010100 || 111110001111110 ... 111101010010111 	   3	  17	-0.3603100204852186
111110110010001 ... 111110110010001 || 100101000011100 ... 111000110100010 	   9	  10	-0.254079327688995
001010010000000 ... 001010010000000 || 100111101001111 ... 011111000011110 	   3	  15	0.3651577081778396

tools._read_chromosome

This function read the chromosome information from a file and it is loaded in memory. The file can be generated by to_json() method.

The docstring associated to the _read_chromosome function is the following.

?_read_chromosome

Parameters
----------
path_or_buf :  string or file handle. File path or object.

Returns
-------
This function returns a chromosome

Examples
--------
>>> A = chromosome(genes = '00100101001001011000111010110010', nko=16, nwt=16, fitness=0.98)
>>> A.print(5)
00100 ... 00101 || 10001 ... 10010     6       8               0.98
>>> A.to_json('prueba.json')
>>> B = _read_chromosome('prueba.json')
>>> B.print(10)
0010010100 ... 0100100101 || 1000111010 ... 1010110010         6       8               0.98

new_CH = _read_chromosome(path_or_buf = 'out.json')
new_CH.print()

0110101101 ... 0110101101 || 011000111101101 ... 011010110001111 	   6	  18	         0.4125

# Examples

Import the pygenmet package

from pygenmet import *

import pandas as pd
import numpy as np
import random
import datetime

We define a new fitness function.

Definition of a simple fitness function

We define a simple fitness function. This function always returns the 1.0 value.

def my_own_get_fitness(genes, nwt, nko, dwt, dko, obj):

    p = 1.0
   
    return p

Now, we generate 10 initial solutions (by _get_father_function) and we can see the fitness function for each solution.

nwt = 15
nko = 25
dwt = pd.DataFrame(np.array(random.choices('01', k = nwt * 10)).reshape(nwt, 10))
dko = pd.DataFrame(np.array(random.choices('01', k = nko * 10)).reshape(nko, 10))

for i in np.arange(10):
    A = _get_father(nwt = nwt, nko = nko, dwt = dwt, dko = dko, 
                    geneSet = '01', get_fitness = my_own_get_fitness, 
                    obj = 1)

    A.print()

110011000001000 ... 110011000001000 || 110011001111101 ... 111011110111010 	   5	  17	            1.0
111110000101011 ... 111110000101011 || 100110000001111 ... 011110100001011 	   9	  11	            1.0
010001110001001 ... 010001110001001 || 101010000001001 ... 010011011011110 	   6	  12	            1.0
100110001011100 ... 100110001011100 || 011111100111111 ... 111111101001100 	   7	  17	            1.0
111101110100010 ... 111101110100010 || 001000001001111 ... 011111011000001 	   9	  10	            1.0
001110011101011 ... 001110011101011 || 001110001101100 ... 011001000001110 	   9	  11	            1.0
100110010100001 ... 100110010100001 || 101011001110011 ... 100111110100001 	   6	  14	            1.0
101000101010110 ... 101000101010110 || 011110001000010 ... 000101001101101 	   7	  12	            1.0
101101000111111 ... 101101000111111 || 010001011010110 ... 101100011110110 	  10	  13	            1.0
100010011101000 ... 100010011101000 || 100000001100100 ... 001000101000100 	   6	   7	            1.0

Definition of a little more complex function

In this case, we considerer the following fitness function.

def my_own_get_fitness(genes, nwt, nko, dwt, dko, obj):
    
    N_WT = sum(1 for x in genes[:nwt] if x == '1')
    N_KO = sum(1 for x in genes[nwt:] if x == '1')

    p = N_WT / N_KO
    
    return p

nwt = 15
nko = 25
dwt = pd.DataFrame(np.array(random.choices('01', k = nwt * 10)).reshape(nwt, 10))
dko = pd.DataFrame(np.array(random.choices('01', k = nko * 10)).reshape(nko, 10))

for i in np.arange(10):
    A = _get_father(nwt = nwt, nko = nko, dwt = dwt, dko = dko, 
                    geneSet = '01', get_fitness = my_own_get_fitness, 
                    obj = 1)

    A.print()

101101010001100 ... 101101010001100 || 001111100111000 ... 110001000000010 	   7	  10	            0.7
011111101110001 ... 011111101110001 || 101011110010001 ... 100010110010100 	  10	  12	0.8333333333333334
100010101100111 ... 100010101100111 || 101011111110000 ... 100000101111010 	   8	  15	0.5333333333333333
111111010110110 ... 111111010110110 || 100010110011100 ... 111001100100110 	  11	  12	0.9166666666666666
110001100110000 ... 110001100110000 || 000001010000100 ... 001001111000011 	   6	   9	0.6666666666666666
000110110000011 ... 000110110000011 || 001111111111110 ... 111100011000111 	   6	  17	0.35294117647058826
011100100101000 ... 011100100101000 || 001000100101111 ... 011110110100010 	   6	  11	0.5454545454545454
100111100010001 ... 100111100010001 || 101110000110001 ... 100011110010011 	   7	  13	0.5384615384615384
101101100110001 ... 101101100110001 || 010100101010101 ... 101010110100011 	   8	  12	0.6666666666666666
111110010111010 ... 111110010111010 || 100011110101101 ... 011011010000010 	  10	  12	0.8333333333333334

It is easy to calculate the fitness function. For example, in the first solution the fitness value is 7/10 = 0.7.

Using a fitness function with mutation operation to find an optimum

In this example, the optimum will be find when the genes string will be '111111...11111'. In this case, p = 0

def my_own_get_fitness(genes, nwt, nko, dwt, dko, obj):
    
    N_WT = sum(1 for x in genes[:nwt] if x == '1') - nwt
    N_KO = sum(1 for x in genes[nwt:] if x == '1') - nko

    p = N_WT + N_KO
    
    return p

nwt = 15
nko = 25
dwt = pd.DataFrame(np.array(random.choices('01', k = nwt * 10)).reshape(nwt, 10))
dko = pd.DataFrame(np.array(random.choices('01', k = nko * 10)).reshape(nko, 10))

for i in np.arange(10):
    A = _get_father(nwt = nwt, nko = nko, dwt = dwt, dko = dko, 
                    geneSet = '01', get_fitness = my_own_get_fitness, 
                    obj = 0)

    A.print()

101000001010000 ... 101000001010000 || 000001010101001 ... 010010010010010 	   4	   8	            -28
010001000011010 ... 010001000011010 || 111100010001000 ... 010001111001000 	   5	  11	            -24
110011100110111 ... 110011100110111 || 011111111100110 ... 001100110110011 	  10	  17	            -13
100111010100011 ... 100111010100011 || 001011100010000 ... 100001011100010 	   8	  10	            -22
110001110110110 ... 110001110110110 || 110011011101011 ... 010110111100110 	   9	  16	            -15
111000100101110 ... 111000100101110 || 011101011110100 ... 101001011011001 	   8	  15	            -17
100010000110110 ... 100010000110110 || 010010011000000 ... 000001011111001 	   6	  11	            -23
001110011110111 ... 001110011110111 || 100011011100010 ... 000101010110111 	  10	  14	            -16
101010101010011 ... 101010101010011 || 011110111100010 ... 000101100001100 	   8	  13	            -19
000001110111111 ... 000001110111111 || 011010000011111 ... 111110000000001 	   9	   9	            -22

We initialize the algorithm with a maximum of 1000 generations. When the iterations reached the 1000 iterations, the algorithm stops and returns the optimum reached.

Moreover, in the following example, the _mutationfunction is applied.

generations = 1000
starttime = datetime.datetime.now()

nwt = 25
nko = 50
dwt = pd.DataFrame(np.array(random.choices('01', k = nwt * 20)).reshape(nwt, 20))
dko = pd.DataFrame(np.array(random.choices('01', k = nko * 20)).reshape(nko, 20))

A = _get_father(nwt = nwt, nko = nko, dwt = dwt, dko = dko, 
                geneSet = '01', get_fitness = my_own_get_fitness, 
                obj = 0)

for i in np.arange(generations):
    A_son = _mutation(A.Genes, geneSet = '01', 
                      get_fitness = my_own_get_fitness, 
                      nwt = nwt, nko = nko, dwt = dwt, dko = dko, 
                      obj = 0, type = 'bsm', p = None)
    
    if A_son.Fitness > A.Fitness:
        A = A_son
        _show_partial_solution(A, starttime = starttime, N = 4)

0101...1111 || 1010...0001 	  14	  22	                  -39	 time:    0.001
0101...1111 || 1110...0001 	  16	  24	                  -35	 time:    0.002
0101...1111 || 1110...0001 	  16	  25	                  -34	 time:    0.002
0101...1111 || 1110...0001 	  16	  26	                  -33	 time:    0.003
0101...1111 || 1110...0001 	  17	  27	                  -31	 time:    0.003
0101...1111 || 1110...0001 	  17	  28	                  -30	 time:    0.003
0101...1111 || 1110...0001 	  17	  29	                  -29	 time:    0.003
0101...1111 || 1110...0001 	  18	  29	                  -28	 time:    0.003
0101...1111 || 1110...0001 	  18	  30	                  -27	 time:    0.003
0101...1111 || 1110...0001 	  18	  31	                  -26	 time:    0.004
0101...1111 || 1110...0001 	  18	  32	                  -25	 time:    0.004
0101...1111 || 1110...0001 	  19	  32	                  -24	 time:    0.007
0101...1111 || 1110...0001 	  19	  33	                  -23	 time:    0.008
0101...1111 || 1110...0001 	  19	  34	                  -22	 time:    0.008
1101...1111 || 1110...0001 	  20	  34	                  -21	 time:    0.009
1101...1111 || 1110...0001 	  20	  35	                  -20	 time:    0.009
1101...1111 || 1110...0001 	  20	  36	                  -19	 time:    0.010
1101...1111 || 1110...0001 	  20	  37	                  -18	 time:    0.011
1101...1111 || 1110...0001 	  20	  38	                  -17	 time:    0.011
1101...1111 || 1110...0001 	  20	  39	                  -16	 time:    0.011
1101...1111 || 1110...0001 	  21	  39	                  -15	 time:    0.011
1101...1111 || 1110...0001 	  22	  39	                  -14	 time:    0.011
1101...1111 || 1110...0001 	  23	  39	                  -13	 time:    0.012
1101...1111 || 1110...0001 	  24	  39	                  -12	 time:    0.013
1101...1111 || 1110...0001 	  24	  40	                  -11	 time:    0.017
1101...1111 || 1110...0011 	  24	  41	                  -10	 time:    0.018
1101...1111 || 1111...0011 	  24	  42	                   -9	 time:    0.024
1111...1111 || 1111...0011 	  25	  42	                   -8	 time:    0.025
1111...1111 || 1111...0011 	  25	  43	                   -7	 time:    0.026
1111...1111 || 1111...0011 	  25	  44	                   -6	 time:    0.032
1111...1111 || 1111...0011 	  25	  45	                   -5	 time:    0.032
1111...1111 || 1111...1011 	  25	  46	                   -4	 time:    0.036
1111...1111 || 1111...1111 	  25	  47	                   -3	 time:    0.040
1111...1111 || 1111...1111 	  25	  48	                   -2	 time:    0.050
1111...1111 || 1111...1111 	  25	  49	                   -1	 time:    0.056
1111...1111 || 1111...1111 	  25	  50	                    0	 time:    0.056

We can see how the algorithm reaches the optimum at 0.056 seconds.

import matplotlib.pyplot as plt

fig, ax = plt.subplots(nrows = 1, ncols = 1, figsize = (10, 6))

generations = 10000
starttime = datetime.datetime.now()

nwt = 250
nko = 500
dwt = pd.DataFrame(np.array(random.choices('01', k = nwt * 50)).reshape(nwt, 50))
dko = pd.DataFrame(np.array(random.choices('01', k = nko * 50)).reshape(nko, 50))
 
for k in [1,2,3,4,5]:
    
    p = 10**(-k)

    results = []
    
    starttime = datetime.datetime.now()
    
    A = _get_father(nwt = nwt, nko = nko, dwt = dwt, dko = dko, 
                    geneSet = '01', get_fitness = my_own_get_fitness, 
                    obj = 0, random_state = 42)

    for i in np.arange(generations):
        A_son = _mutation(A.Genes, geneSet = '01', 
                          get_fitness = my_own_get_fitness, 
                          nwt = nwt, nko = nko, dwt = dwt, dko = dko, 
                          obj = 0, type = 'bsm', p = p)

        if A_son.Fitness > A.Fitness:
            A = A_son
            results.append([p, A.Fitness, (datetime.datetime.now() - starttime).total_seconds()])            
      
    results = pd.DataFrame(results, columns = ['p', 'fitness', 'time'])
    ax.plot(results.time, results.fitness, label = r'p = $10^{-{' + str(k) + '}}$', linewidth = 2)

ax.grid(linestyle = '--', color = 'gray', alpha = 0.5)
ax.set_title(r'Convergence of the genetic algorithm by mutation probability (10000 generations)')
ax.set_ylabel(r'Fitness value')
ax.set_xlabel(r'Time (seconds)')
ax.legend(frameon = False, title = r'Probability', bbox_to_anchor = [1.0, 1.0]);

References

Alonso, C., Fernández-Ramos, D., Varela-Rey, M., Martínez-Arranz, I., Navasa, N., Van Liempd, S. M., Lavin, J. L., Mayo, R., Ilisso, C. P., de Juan, V. G., Iruarrizaga-Lejarreta, M., delaCruz-Villar, L., Mincholé, I., Robinson, A., Crespo, J., Martín-Duce, A., Romero-Gomez, M., Sann, H., Platon, J., ... Mato, J. M. (2017). Metabolomic Identification of Subtypes of Nonalcoholic Steatohepatitis. Gastroenterology. https://doi.org/10.1053/j.gastro.2017.01.015

Iruarrizaga-Lejarreta, M., Varela-Rey, M., Fernández-Ramos, D., Martínez-Arranz, I., Delgado, T. C., Simon, J., Gutiérrez-de Juan, V., delaCruz-Villar, L., Azkargorta, M., Lavin, J. L., Mayo, R., Van Liempd, S. M., Aurrekoetxea, I., Buqué, X., Delle Cave, D., Peña, A., Rodríguez-Cuesta, J., Aransay, A. M., Elortza, F., ... Mato, J. M. (2017). Role of aramchol in steatohepatitis and fibrosis in mice. Hepatology Communications. https://doi.org/10.1002/hep4.1107

Martinez-Arranz, I., Mayo, R., Pérez-Cormenzana, M., Mincholé, I., Salazar, L., Alonso, C., & Mato, J. M. (2015). Enhancing metabolomics research through data mining. J Proteomics, 127(Pt B), 275–288. https://doi.org/10.1016/j.jprot.2015.01.019

Morrison, M. C., Verschuren, L., Salic, K., Verheij, J., Menke, A., Wielinga, P. Y., Iruarrizaga-Lejarreta, M., Gole, L., Yu, W.-M., Turner, S., Caspers, M. P. M., Martínez-Arranz, I., Pieterman, E., Stoop, R., van Koppen, A., van den Hoek, A. M., Mato, J. M., Hanemaaijer, R., Alonso, C., & Kleemann, R. (2018). Obeticholic Acid Modulates Serum Metabolites and Gene Signatures Characteristic of Human NASH and Attenuates Inflammation and Fibrosis Progression in Ldlr-/-.Leiden Mice. Hepatology Communications, 2(12), 1513–1532. https://doi.org/10.1002/hep4.1270

Affenzeller, M., Wagner, S., Winkler, S., & Beham, A. (2009). Genetic Algorithms and Genetic Programming: Modern Concepts and Practical Applications. CRC Press.

Cavill, R., Keun, H. C., Holmes, E., Lindon, J. C., Nicholson, J. K., & Ebbels, T. M. D. (2009). Genetic algorithms for simultaneous variable and sample selection in metabonomics. Bioinformatics (Oxford, England), 25(1), 112–118. https://doi.org/10.1093/bioinformatics/btn586

Chambers, L. (1995). Practical Handbook of Genetic Algorithms: Volume 1: Applications. CRC-Press.

Chambers, L. D. (2019a). Practical Handbook of Genetic Algorithms: Complex Coding Systems, Volume III. CRC Press.

Chambers, L. D. (2019b). The Practical Handbook of Genetic Algorithms: New Frontiers, Volume II. CRC Press.

Eiben, A. E., & Smith, J. E. (2015). Introduction to Evolutionary Computing. Springer. Gandomi, A. H., Alavi, A. H., & Ryan, C. (2015). Handbook of Genetic Programming Applications. Springer.

Gen, M., & Cheng, R. (1999). Genetic Algorithms and Engineering Optimization. John Wiley & Sons.

Haupt, R. L., & Haupt, S. E. (2004). Practical Genetic Algorithms. John Wiley & Sons.

Haupt, R. L., & Werner, D. H. (2007). Genetic Algorithms in Electromagnetics. John Wiley & Sons.

Holland, J. H. (1992). Genetic Algorithms. Scientific American, 267(1), 66–73. JSTOR.

Kramer, O. (2017). Genetic Algorithm Essentials. Springer.

Luque, G., & Alba, E. (2011). Parallel Genetic Algorithms: Theory and Real World Applications. Springer Science & Business Media.

Michalewicz, Z. (2013). Genetic Algorithms + Data Structures = Evolution Programs. Springer Science & Business Media.

Mitchell, M. (1998). An Introduction to Genetic Algorithms. MIT Press.

Petrowski, A., & Ben-Hamida, S. (2017). Evolutionary Algorithms. John Wiley & Sons.

Reeves, C. R., & Rowe, J. E. (2006). Genetic Algorithms: Principles and Perspectives: A Guide to GA Theory. Springer Science & Business Media.

Sheppard, C. (2018). Genetic Algorithms with Python. Clinton Sheppard.

Shiffman, D. (2012). The Nature of Code. D. Shiffman.

Sivanandam, S. N., & Deepa, S. N. (2007). Introduction to Genetic Algorithms. Springer Science & Business Media.

Walchand College of Engineering, A.J., U., P.D., S., & Government College of Engineering, Karad. (2015). CROSSOVER OPERATORS IN GENETIC ALGORITHMS: A REVIEW. ICTACT Journal on Soft Computing, 06(01), 1083–1092. https://doi.org/10.21917/ijsc.2015.0150

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

README.md

README.md

How to use

Introduction

CLASS `chromosome`

Methods

print()

to_json()

fitness

fitness._get_fitness

fitness._get_fitness_robust

fitness._get_fitness_l

operators

Crossover

operators._crossover

operators._arithmetic_crossover

Mutation

operators._mutation

Selection

operators._replacement

operators._selection

tools

tools._get_father

tools._get_population

tools._read_chromosome

Definition of a simple fitness function

Definition of a little more complex function

Using a fitness function with mutation operation to find an optimum

References

Files

README.md

Latest commit

History

README.md

File metadata and controls

How to use

Introduction

CLASS chromosome

Methods

print()

to_json()

fitness

fitness._get_fitness

fitness._get_fitness_robust

fitness._get_fitness_l

operators

Crossover

operators._crossover

operators._arithmetic_crossover

Mutation

operators._mutation

Selection

operators._replacement

operators._selection

tools

tools._get_father

tools._get_population

tools._read_chromosome

Definition of a simple fitness function

Definition of a little more complex function

Using a fitness function with mutation operation to find an optimum

References

CLASS `chromosome`