reconstruction.ecoli.dataclasses.process.rna_decay

SimulationData for rna decay process

class reconstruction.ecoli.dataclasses.process.rna_decay.RnaDecay(raw_data, sim_data)[source]

Bases: object

_buildRnaDecayData(raw_data, sim_data)[source]
km_loss_function(vMax, rnaConc, kDeg, isEndoRnase, alpha)[source]

Generates the functions used for estimating the per-RNA affinities (Michaelis-Menten constants) to the endoRNAses.

The goal is to find a set of Michaelis-Menten constants such that the endoRNAse-mediated degradation under basal concentrations is consistent with the experimentally observed half-lives.

If nonlinear is the rate of degradation predicted by Michaelis-Menten kinetics and linear is the rate of degradation from observed half-lives, we want:

nonlinear - linear = 0

In reality, there will be residuals R_aux = ``nonlinear - linear. We care about the residuals after normalizing by the linear rate:: R = nonlinear / linear - 1.

In order to turn this into a minimization problem, we define the loss function as the squared sum of the residuals. Additionally, to ensure that all Km values are positive, the loss function accepts as input the logarithm of the final Km values and exponentiates them before calculating the residuals.

The third-party package Autograd uses autodiff to calculate Jacobians for our loss function that can be used during minimization.

Parameters:

  • (float) (vMax)

  • (np.ndarray) (isEndoRnase)

  • (np.ndarray)

  • (np.ndarray)

Returns:

A tuple containing the following functions:

  • L (function): The loss function to minimize.

  • Lp (function): The Jacobian of the loss function L.

  • residual_f (function): The residual error function.

  • residual_aux_f (function): The unnormalized residual error function.

Return type:

tuple