Pattern recognition programming to predict productivity of Yarrowia lipolytica DSM 3286 for citric acid production

Abstract

The non-conventional yeast Yarrowia lipolytica is attracting increasing attention due to its potential to produce large amounts of organic acids from hydrophobic substrates. Due to the steadily increasing demand for citric acid in the industrial sector, the aim of this scientific work was to develop a predictive model of the citric acid productivity of the strain Yarrowia lipolytica DSM3286. As a basis for this, the optical density, pH, cell number and citric acid were determined in 18 identical mixtures.

The citric acid concentration (mean values of the measured concentration over time) follows a linear increase. Based on this, the mathematical calculation operation of linear regression was selected for modeling the prediction model in Python. The following coefficients were determined for the variables used in the learning algorithm:

•       time:                      6,104 * 10^-4

•       OD:                        -1,224 * 10^-1

•       pH value:              -4,043 * 10^-1

•       Cell count:            1,749 * 10^-8

In final validation of the program, a result accuracy of 86.5% was obtained. The result obtained in the present scientific work shows that by means of simple linear regression, over a cultivation period of 13 days, a prediction of the citric acid productivity of strain Yarrowia lipolytica DSM3286 is possible.

Mehr zum Titel