Fed-Batch data generation
Within the simulator, one class is defined as the ‘fedbatch’ (relies on the ‘batch’ class) and it can be used to simulate fed-batch reactors with defined kinetics.
First, the package is loaded and available default examples are displayed using show_implemented_examples method.
[2]:
from insidapy.simulate.ode import fedbatch
fedbatch().show_implemented_examples()
[+] The following examples are implemented in this FEDBATCH class:
+----------------------+----------------------------------------------------------------------------------+
| Example ID string | Description |
+----------------------+----------------------------------------------------------------------------------+
| fedbatch1 | Bioreaction in fedbatch operation mode. Constant input flow. 3 species and 1 |
| | volume. Bacteria growth, substrate consumption and product formation. Mimics the |
| | production of a target protein. |
+----------------------+----------------------------------------------------------------------------------+
Next, one can decide to use one of the implemented examples or define a new one. In this case, the example of a bioreaction in fedbatch operation mode with 3 species is used ('fedbatch1'). Each example would be fully defined with default values. However, we can overwrite these values by using the corresponding inputs. If we overwrite the default values, we are reminded to check that the given values are in correct order.
[3]:
data = fedbatch(example='fedbatch1', # Choose example. Defaults to "fedbatch1".
nbatches=4, # Number of batches. Defaults to 3.
npoints_per_batch=20, # Number of points per batch and per species. Defaults to 20.
noise_mode='percentage', # Noise mode. Defaults to "percentage".
noise_percentage=2.5, # Noise percentage (in case mode is "percentage")
random_seed=10, # Random seed for reproducibility. Defaults to 0.
bounds_initial_conditions=[[0.1, 50, 0.1, 0.1], [0.4, 90, 0.2, 10]], # Bounds for initial conditions. Defaults to "None".
time_span=[0, 80], # Time span for integration. Defaults to "None".
initial_condition_generation_method='LHS', # Method for generating initial conditions. Defaults to "LHS".
name_of_time_vector='time') # Name of time vector. Defaults to "time".
[!] IMPORTANT: It seems that you changed the default bounds of the species. Make sure the order of the indicated bounds is the following: ['biomass', 'product', 'substrate', 'volume']
We can now print some information about the example.
[4]:
data.print_info()
[+] Loaded the example FEDBATCH1 with the following properties:
+--------------------------------+------------------------------------------------------------------------+
| Property | Description |
+--------------------------------+------------------------------------------------------------------------+
| Example string | fedbatch1 |
| Example description | Bioreaction in fedbatch operation mode. Constant input flow. 3 species |
| | and 1 volume. Bacteria growth, substrate consumption and product |
| | formation. Mimics the production of a target protein. |
| Short reference information | ISBN 978-1-119-28591-5 |
| Number of species | 4 |
| Species names | ['biomass', 'product', 'substrate', 'volume'] |
| Species units | ['g/L', 'g/L', 'g/L', 'L'] |
| Number of batches | 4 |
| Number of samples | 20 |
| Time span | [0, 80] |
| Time unit | h |
| Noise mode | percentage |
| Noise percentage | 2.5% |
| Lower bounds for experiments | [0.1, 50.0, 0.1, 0.1] |
| Upper bounds for experiments | [0.4, 90.0, 0.2, 10.0] |
+--------------------------------+------------------------------------------------------------------------+
This might be useful to use the example in a publication (use the short reference information given to reference where the example came from) or to check if the example is defined as intended.
After that, we can run the experiments to create some in-silico data using the run_experiments method. We can then for example check the data of the first experiment.
[5]:
data.run_experiments()
print(data.y_noisy[0])
[+] Experiments done.
time biomass product substrate volume
0 0.000000 0.170610 62.310537 0.195585 3.870119
1 4.210526 0.229802 59.563976 0.564204 4.196437
2 8.421053 0.273714 56.704083 0.799884 4.333703
3 12.631579 0.439917 55.419851 0.943265 4.607631
4 16.842105 0.648348 51.809027 0.885063 4.797903
5 21.052632 0.940617 50.583312 0.680357 5.102327
6 25.263158 1.193427 46.709251 0.446642 5.333964
7 29.473684 1.501493 46.311131 0.261561 5.281053
8 33.684211 1.545675 46.119174 0.183287 5.852505
9 37.894737 1.740762 41.893160 0.158977 5.713684
10 42.105263 1.869093 41.030178 0.134442 6.038717
11 46.315789 1.907932 38.812825 0.128943 6.506188
12 50.526316 1.990550 38.634757 0.102781 6.468446
13 54.736842 2.162627 36.016796 0.082139 6.957505
14 58.947368 2.271315 36.663173 0.059360 6.917161
15 63.157895 2.314698 36.818400 0.105836 7.327268
16 67.368421 2.406487 32.865182 0.073221 7.497049
17 71.578947 2.508249 33.826766 0.075424 7.515382
18 75.789474 2.528988 30.561737 0.064671 7.815977
19 80.000000 2.610140 31.081639 0.081498 8.283423
Most modeling approaches require a training dataset and a separate testing dataset. To generate separate datasets, the user can apply a splitting in an sklearn-manner. There is no default value set. In case the user calls the function, a test_splitratio in the range [0,1) needs to be chosen. The value represents the fraction of the total number of batches generated used for the test set. The data is then splitted and stored in the data object as data.training and data.testing.
[6]:
data.train_test_split(test_splitratio=0.2)
We can now plot the experiments. The fist way to do so is to plot all the runs using the plot_experiments method. The method lets us save the figure using a path (save_figure_directory), a name (figname) and an some extensions (save_figure_exensions) as a list. By using show=False, the plot will not be displayed in a running code.
[7]:
data.plot_experiments( save=True,
show= False,
figname=f'{data.example}_simulated_batches',
save_figure_directory=r'.\figures',
save_figure_exensions=['png'])
[+] Saving figure:
->png: .\figures\fedbatch1_simulated_batches.png
We can also visualize the training and testing runs individually using the plot_train_test_experiments method.
[8]:
data.plot_train_test_experiments( save=True,
show=False,
figname=f'{data.example}_simulated_batches_train_test',
save_figure_directory=r'.\figures',
save_figure_exensions=['png'])
[+] Saving figure:
->png: .\figures\fedbatch1_simulated_batches_train_test.png
After the simulation, one can export the data as XLSX files. By choosing which_dataset to be training (only executable if train_test_split was applied), testing (only executable if train_test_split was applied), or all (always executable), the corresponding data is exported to the indicated location:
[9]:
data.export_dict_data_to_excel(destination=r'.\data', which_dataset='all') # Exports all the data
data.export_dict_data_to_excel(destination=r'.\data', which_dataset='training') # Exports the training data (blue circles in Fig 2)
data.export_dict_data_to_excel(destination=r'.\data', which_dataset='testing') # Exports the training data (red diamonds in Fig 2)
[+] Exported batch data to excel.
-> Dataset: ALL (options: training, testing, all)
-> Noise free data to: .\data\fedbatch_fedbatch1_all_batchdata.xlsx
-> Noisy data to: .\data\fedbatch_fedbatch1_all_batchdata_noisy.xlsx
[+] Exported batch data to excel.
-> Dataset: TRAINING (options: training, testing, all)
-> Noise free data to: .\data\fedbatch_fedbatch1_training_batchdata.xlsx
-> Noisy data to: .\data\fedbatch_fedbatch1_training_batchdata_noisy.xlsx
[+] Exported batch data to excel.
-> Dataset: TESTING (options: training, testing, all)
-> Noise free data to: .\data\fedbatch_fedbatch1_testing_batchdata.xlsx
-> Noisy data to: .\data\fedbatch_fedbatch1_testing_batchdata_noisy.xlsx