gaBinaryT1: Genetic algorithm used to optimise a model

In saezlab/CellNOptR: Training of boolean logic models of signalling networks using prior knowledge networks and perturbation data

Genetic algorithm used to optimise a model

Description

This function is the genetic algorithm to be used to optimise a model by fitting to data containing one time point.

Usage

gaBinaryT1(CNOlist, model, initBstring=NULL, sizeFac = 1e-04, 
    NAFac = 1, popSize = 50, pMutation = 0.5, maxTime = 60, maxGens = 500,
    stallGenMax = 100, selPress = 1.2, elitism = 5, relTol = 0.1, verbose=TRUE,
    priorBitString=NULL, timeIndex=2)

Arguments

CNOlist	a CNOlist on which the score is based (based on valueSignals[[2]], i.e. data at time 1)
model	a model structure, as created by readSIF, normally pre-processed but that is not a requirement of this function
initBstring	an initial bitstring to be tested, should be of the same size as the number of reactions in the model above (model$reacID). Default is all ones.
sizeFac	the scaling factor for the size term in the objective function, default to 0.0001
NAFac	the scaling factor for the NA term in the objective function, default to 1
popSize	the population size for the genetic algorithm, default set to 50
pMutation	the mutation probability for the genetic algorithm, default set to 0.5
maxTime	the maximum optimisation time in seconds, default set to 60
maxGens	the maximum number of generations in the genetic algorithm, default set to 500
stallGenMax	the maximum number of stall generations in the genetic algorithm, default to 100
selPress	the selective pressure in the genetic algorithm, default set to 1.2
elitism	the number of best individuals that are propagated to the next generation in the genetic algorithm, default set to 5
relTol	the relative tolerance for the best bitstring reported by the genetic algorithm, i.e., how different from the best solution, default set to 0.1
verbose	logical (default to TRUE) do you want the statistics of each generation to be printed on the screen?
priorBitString	At each generation, the GA algorithm creates a population of bitstrings that will be used to perform the optimisation. If the user knows the values of some bits, they can be used to overwrite bit values proposed by the GA algorithm. If provided, the priorBitString must have the same length as the initial bitstring and be made of 0, 1 or NA (by default, this bitstring is set to NULL, which is equivalent to setting all bits to NA). Bits that are set to 0 or 1 are used to replace the bits created by the GA itself (see example).
timeIndex	the index of the time point to optimize. Must be greater or equal to 2 (1 corresponds to time=0). Must be less than the number of time points. Default is 2.

Details

The whole procedure is described in details in Saez-Rodriguez et al. (2009). The basic principle is that at each generation, the algorithm evaluates a population of models based on excluding or including some gates in the initial pre-processed model (this is encoded in a bitstring with contains 0/1 entries for each gate). The population is then evolved based on the results of the evaluation of these networks, where the evaluation is obtained by simulating the model (to steady state) under the various conditions present in the data, and then computing the squared deviation from the data, to which a penalty is added for size of the model and for species in the model that do not reach steady state.

Value

This function returns a list with elements:

bString	the best bitstring
bScore	the best score
results	a matrix with columns "Generation","Best_score","Best_bitString","Stall_Generation","Avg_Score_Gen","Best_score_Gen","Best_bit_Gen","Iter_time"
stringsTol	the bitstrings whose scores are within the tolerance
stringsTolScores	the scores of the above-mentioned strings

Author(s)

C. Terfve. T. Cokelaer

References

J. Saez-Rodriguez, L. G. Alexopoulos, J. Epperlein, R. Samaga, D. A. Lauffenburger, S. Klamt and P. K. Sorger. Discrete logic modeling as a means to link protein signaling networks with functional analysis of mammalian signal transduction, Molecular Systems Biology, 5:331, 2009.

See Also

gaBinaryTN, simulatorT1

Examples

data(CNOlistToy,package="CellNOptR")
data(ToyModel,package="CellNOptR")

#pre-process model

model = preprocessing(CNOlistToy, ToyModel)

#optimise

initBstring<-rep(1,length(model$reacID))
ToyT1opt<-gaBinaryT1(
	CNOlist=CNOlistToy,
	model=model,
	initBstring=initBstring,
	maxGens=100, popSize=10, verbose=FALSE)

# During the optimisation, some bits can be overwritten by your prior knowledge 
# First, you need to create a priorBitString made of NA where known bit values
# are replaced by 0 or 1
priorBitString = rep(NA, length(model$reacID))
priorBitString[1] = 0
priorBitString[2] = 1

# Second, you call the gaBinaryT1 function by providing the priorBitString
# argument:
ToyT1opt<-gaBinaryT1(CNOlist=CNOlistToy, model=model,
     initBstring=initBstring,	maxGens=10, popSize=10, verbose=FALSE, 
    priorBitString=priorBitString)

saezlab/CellNOptR documentation built on April 16, 2024, 5:21 a.m.