oRes: Import and combine optimisation results

In saezlab/PHONEMeS-ILP: Building logic-based models from discovery high-content phosphoproteomics data

Import and combine optimisation results

Description

Function that imports and combines the results of an optimisation.

Usage

oRes(data.On, optParam, pknList)

Arguments

optParam	an OptParam object with at least elements resN (the index of the optimisation, this is the suffix of the Resultsx.RData files, and matches the suffix of the results folder wd/Results_resN where all results will be written, tol (the tolerance that was used for the optimisation, a number between 0 and 1 if relative tolerance, or a number larger or equal to 1 if absolute tolerance), nG (the number of generations that the optimisation was run for)
data.On	a GMMbyCond object, as created by dataBycond (as it was fed into the optimisation)
pknList	a PKNlist object as it was fed in the optimisation process

Details

This function imports all the output and results files from the optimisation process and returns summarised objects that contain the evolution of scores, weights, etc.

Value

This function returns a list of list:

nM	vector, the number of models that pass the tolerance threshold, at each generation
sM	matrix, score of the best model of the generation, by drug (rows=generations, cols=drugs, in the same order as in data.On)
sM.avg	matrix, average score of the generation, by drug (rows=generations, cols=drugs, in the same order as in data.On)
sAll	matrix, total score of each model, at each generation (rows=generations, cols=models in population, fixed to 5000 currently)
Msize	matrix, size of all models, at each generation, mapping to sAll
BMGsize	vector, size of the best model of each generation
BM.s	vector, total score of the best model of the generation (with size penalty if any)
FE	matrix, counts of each edge at each generation (rows=edges (in the same order as in c.I.list$complete.I), cols=generations); to get the frequency from these numbers, each entry should be divided by the number of models in the population, currently fixed to 5000
intgNone	matrix, counts of "no inputs" for each integrator at every generation (rows=integrator nodes (in the same order as in c.I.list$integrators), cols=generations); to get the frequency from these numbers, each entry should be divided by the number of models in the population, currently fixed to 5000
intgAnd	matrix, counts of "ands" for each integrator at every generation (rows=integrator nodes (in the same order as in c.I.list$integrators), cols=generations); to get the frequency from these numbers, each entry should be divided by the number of models in the population, currently fixed to 5000
G1.freq	matrix, ntag of each edge at each generation (rows=edges (in the same order as in c.I.list$complete.I), cols=generations); these are the number of bins for each edge used for sampling at the next generation
G1.flipP	matrix, probability of "and" for each intermediate at every generation (rows=intermediate nodes (in the same order as in c.I.list$intermediates), cols=generations)
G1.AndBinF	matrix, ntag of the "AND" integrator hyperedge at each generation (rows=integrator nodes (in the same order as in c.I.list$integrators), cols=generations); these are the number of bins for the AND hyperedge for each integrator node, used for sampling at the next generation
G1.NoneBinF	matrix, ntag of the "none" integrator hyperedge at each generation (rows=integrator nodes (in the same order as in c.I.list$integrators), cols=generations); these are the number of bins for the "no input" edge for each integrator node, used for sampling at the next generation

Author(s)

C. Terfve

saezlab/PHONEMeS-ILP documentation built on June 21, 2022, 5:36 p.m.