Skip to main content

Table 3 Results for Case Study II

From: Modelling biochemical networks with intrinsic time delays: a hybrid semi-parametric approach

NN

Nlag

τ

BIC

  

MSE

  

NN

Nlag

τ

BIC

  

MSE

  
   

train

valid

test

train

valid

test

   

train

valid

test

train

valid

test

3

0

0

-18561

-5147

-5000

0.0155

0.0187

0.0157

6

0

0

-18500

-5199

-5030

0.0148

0.0187

0.0154

4

0

0

-18514

-5164

-4994

0.0155

0.0187

0.0150

7

0

0

-18576

-5197

-4990

0.0153

0.0180

0.0144

5

0

0

-18531

-5186

-5011

0.0151

0.0189

0.0153

8

0

0

-18430

-5211

-5053

0.0146

0.0182

0.0158

2

1

2

-17144

-4697

-4475

0.0086

0.0098

0.0065

2

1

2.5

-16343

-4293

-4331

0.0077

0.0076

0.0067

3

1

2

-16981

-4588

-4470

0.0085

0.0108

0.0067

3

1

2.5

-17230

-4725

-4584

0.0186

0.0327

0.0183

4

1

2

-16877

-4581

-4484

0.0087

0.0117

0.0072

4

1

2.5

-16635

-4506

-4690

0.0154

0.0220

0.0170

5

1

2

-16792

-4544

-4490

0.0078

0.0093

0.0064

5

1

2.5

-16927

-4614

-4524

0.0086

0.0111

0.0072

6

1

2

-16911

-4643

-4544

0.0087

0.0119

0.0073

6

1

2.5

-16944

-4523

-4488

0.0082

0.0092

0.0066

7

1

2

-17197

-4632

-4505

0.0097

0.0151

0.0084

7

1

2.5

-17044

-4742

-4608

0.0082

0.0107

0.0075

8

1

2

-17181

-4530

-4496

0.0123

0.0155

0.0105

8

1

2.5

-16976

-4694

-4608

0.0084

0.0110

0.0076

2

2

2

-16466

-4512

-4525

0.0067

0.0079

0.0067

2

2

2.5

-17813

-4833

-4759

0.0130

0.0152

0.0129

3

2

2

-16856

-4656

-4535

0.0081

0.0116

0.0073

3

2

2.5

-16637

-4684

-4595

0.0079

0.0123

0.0079

4

2

2

-16788

-4616

-4525

0.0079

0.0111

0.0072

4

2

2.5

-16703

-4507

-4517

0.0073

0.0083

0.0064

5

2

2

-16734

-4430

-4446

0.0075

0.0088

0.0061

5

2

2.5

-16384

-4327

-4404

0.0068

0.0074

0.0063

6

2

2

-16573

-4271

-4353

0.0077

0.0081

0.0065

6

2

2.5

-16601

-4400

-4432

0.0071

0.0079

0.0062

7

2

2

-16704

-4569

-4541

0.0071

0.0089

0.0065

7

2

2.5

-16569

-4405

-4466

0.0068

0.0072

0.0061

8

2

2

-16921

-4728

-4632

0.0082

0.0132

0.0079

8

2

2.5

-16833

-4790

-4664

0.0080

0.0127

0.0079

2

3

2

-19006

-5136

-5080

0.0181

0.0177

0.0158

2

3

2.5

-16619

-4566

-4514

0.0077

0.0099

0.0071

3

3

2

-16811

-4692

-4549

0.0078

0.0119

0.0073

3

3

2.5

-16037

-4218

-4259

0.0064

0.0072

0.0058

4

3

2

-16737

-4474

-4466

0.0069

0.0089

0.0063

4

3

2.5

-16439

-4224

-4287

0.0068

0.0078

0.0057

5

3

2

-16519

-4357

-4408

0.0066

0.0076

0.0058

5

3

2.5

-16199

-4358

-4295

0.0063

0.0076

0.0056

6

3

2

-16832

-4506

-4415

0.0090

0.0107

0.0078

6

3

2.5

-16604

-4577

-4556

0.0072

0.0094

0.0067

7

3

2

-16565

-4385

-4439

0.0066

0.0072

0.0058

7

3

2.5

-16344

-4475

-4432

0.0064

0.0078

0.0060

8

3

2

-16758

-4672

-4569

0.0079

0.0117

0.0069

8

3

2.5

-16471

-4374

-4505

0.0066

0.0069

0.0061

2

4

2

-16655

-4365

-4504

0.0071

0.0107

0.0077

2

4

2.5

-16562

-4532

-4503

0.0079

0.0097

0.0073

3

4

2

-16377

-4301

-4431

0.0067

0.0078

0.0064

3

4

2.5

-16325

-4471

-4470

0.0072

0.0086

0.0066

4

4

2

-16215

-4183

-4316

0.0062

0.0067

0.0057

4

4

2.5

-16261

-4189

-4281

0.0064

0.0068

0.0058

5

4

2

-16611

-4481

-4484

0.0070

0.0101

0.0066

5

4

2.5

-15954

-4190

-4255

0.0056

0.0060

0.0052

6

4

2

-17503

-4762

-4597

0.0189

0.0374

0.0197

6

4

2.5

-16439

-4334

-4476

0.0064

0.0073

0.0060

7

4

2

-25835

-7171

-7226

10.160

9.2368

11.843

7

4

2.5

-25949

-7288

-7280

13.781

19.040

17.824

8

4

2

-16256

-4328

-4369

0.0058

0.0061

0.0052

8

4

2.5

-16296

-4326

-4381

0.0061

0.0065

0.0054

  1. Results of the performance measures, BIC and MSE, over structure parameters, namely Numbers of Nodes in the hidden layer of the ANN, NN, Number of time lags, Nlag and the time lag, τ, for Pichia pastoris cells with distributed time delays using the structure of Fig. 4B. Integration of material balances along with the equations obtained from the sensitivity method is carried out using the linear approximation integration schema described in the Methods section. The times series were chosen such that one of the delays matched the maximum of the time delay of the weighting function of the simulation case (see Eqs. A 12).