Modeling the Biocatalytic Method of Lipid Extraction Using Artificial Neural Networks

Round 1

Reviewer 1 Report

Comment to the authors

I proceeded to analyze the manuscript entitled:

Modeling the biocatalytic method of lipid extraction using artificial neural networks

written by:

Anton V. Shafrai, Alexander Yu. Prosekov, Elena A. Vechtomova

The manuscript describes the result of a study carried on to apply an ANN for mathematical modeling of the biocatalytic method for obtaining rendered game animal fat. The study presents the design to an ANN for predicting the output values of the process, the evaluating the ANN accuracy and the process optimization. The experiment was carried out on the fat of bear, beaver, marmot and badger hunted in Kemerovo Region, Siberia, Russia.

The topic is, in my opinion, interesting and using ANN in optimization is quite actual.

The experimental setup is ingenious and elaborated. The figures are suggestive and support the statements. References are in proper amount and indicate that the authors are well aware of what has been published on the subject they are writing about. The article is well written, using good English, in my opinion. The ANN was trained on experimental data, which involved a considerable effort to acquire them.

Moving to details, I found a few parts that, in my opinion, require a slight improvement and additional clarification, and they are mentioned below.

-The introduction is comprehensive with respect to the biocatalytic method for obtaining rendered animal fat but mentions just a long list of references. Rewrite that mentioning some details on the references on the ANN part of the Introduction.

-Page 2 you wrote:”Prepared samples of raw fat amounting to 50 g were mixed with water at 1:1.” Clarify in the manuscript if is volume or mass ratio.

-Page4: “Figure 2. Change in the yield of rendered fat with the Neutrase enzyme preparation” Mention in the manuscript what type of was used. The same for Figure 5.

-Page 6: “Out of the147 dataset records, 98 were put in the training set (67%), and 49 were put in the test set (33%).” Did you use a validation set or not? Mention this in your manuscript.

Page7: “The ANN comprised 8 hidden layers with 64 neurons in each.” Explain in your manuscript how you decided for this ANN architecture. It is quite a big number of neurons yo uused. Is there such a big number of layers and of neurons in each layer necessary? Explain in your manuscript whether you considered and tried other structures, as well.

Mention the type of computer platform you used and the time required for training.

-Pages 7 and 8: Are the plots in Figures 6 and 7 necessary, once you provided the equations of the functions, as they are not experimental results? If you insist on keeping them, repolt Fig.7 with the Output values between 0 and 1, as the rest is irrelevant.

-An overall observation regarding the Conclusion, only if you believe that this is applicable: Does there male sense to used more than one enzyme type, meaning a mix? If this makes sense, please mention in the manuscript what is the outcome of the ANN for such an input. Maybe it would be interesting to see if a combination of enzymes has a better output than the sum of the outputs for each enzyme used alone, even if the ANN was not trained for such input.

Author Response

Thank you for your comments!

-The introduction is comprehensive with respect to the biocatalytic method for obtaining rendered animal fat but mentions just a long list of references. Rewrite that mentioning some details on the references on the ANN part of the Introduction.

Introduction has been corrected

-Page 2 you wrote:”Prepared samples of raw fat amounting to 50 g were mixed with water at 1:1.” Clarify in the manuscript if is volume or mass ratio.

Prepared samples of raw fat in the amount of 50 g were mixed with water in a ratio of 1:1 by weight

-Page4: “Figure 2. Change in the yield of rendered fat with the Neutrase enzyme preparation” Mention in the manuscript what type of was used. The same for Figure 5.

Fat types added

-Page 6: “Out of the147 dataset records, 98 were put in the training set (67%), and 49 were put in the test set (33%).” Did you use a validation set or not? Mention this in your manuscript.

Validation set was not used

Page7: “The ANN comprised 8 hidden layers with 64 neurons in each.” Explain in your manuscript how you decided for this ANN architecture. It is quite a big number of neurons yo uused. Is there such a big number of layers and of neurons in each layer necessary? Explain in your manuscript whether you considered and tried other structures, as well.

Architecture building algorithm added

Mention the type of computer platform you used and the time required for training.

Type of computer platform and the time required for training added

-Pages 7 and 8: Are the plots in Figures 6 and 7 necessary, once you provided the equations of the functions, as they are not experimental results? If you insist on keeping them, repolt Fig.7 with the Output values between 0 and 1, as the rest is irrelevant.

Figures removed

-An overall observation regarding the Conclusion, only if you believe that this is applicable: Does there male sense to used more than one enzyme type, meaning a mix? If this makes sense, please mention in the manuscript what is the outcome of the ANN for such an input. Maybe it would be interesting to see if a combination of enzymes has a better output than the sum of the outputs for each enzyme used alone, even if the ANN was not trained for such input.

The use of a mixture of enzymes for hydrolysis is possible provided a more detailed study of the structure of protein compounds of raw fat. However, this is difficult when implementing a biocatalytic method for extracting fats of hunting animals on an industrial scale due to the instability of qualitative and quantitative physico-chemical parameters of raw materials, including protein compounds of raw fat. To implement such a method, a large array of experimental data obtained from the study of raw fat isolated from different types of hunting animals, taking into account their age characteristics, habitat, trapping conditions, nutrition, season, etc. is needed.

Reviewer 2 Report

1.      The abstract is of poor quality, although the authors neatly wrote what this article presents. The problem is that the introduction does not contain a clearly described main goal of this work and there is no description of the methods of solving the main task. The summary must necessarily be changed.

2.      The developed graphical and analytical model is correct.

3.      In the summary, the authors write, “Mathematical modeling was carried out using an artificial neural network – a very accurate tool which is popular nowadays”. This sentence is too concise and not precise, please change it.

4.      The caption under Fig. 4 is in bold. Figure 4. Change in the yield of rendered marmot fat with the Fan Boost enzyme preparation

5.      Dependence no. 1 has a wrong notation, if there is normalization, it must be an absolute value, otherwise it does not make mathematical sense

6.      All formulas are written in small font and are hard to read.

7.      In Fig. 10 I don't understand, at 70° there is a low concentration of enzymes, is this correct. If so, why.

8.      in the summary the authors write. “As a result of optimization, the data on the effect of temperature, process duration, introduced enzyme preparation and its concentration on the yield of rendered fat were obtained. there is nothing in the text about it. Please explain.

Author Response

Thank you for your comments!

1. The abstract is of poor quality, although the authors neatly wrote what this article presents.The problem is that the introduction does not contain a clearly described main goal of this work and there is no description of the methods of solving the main task.The summary must necessarily be changed.

Introduction has been corrected

2. The developed graphical and analytical model is correct.

3. In the summary, the authors write, “Mathematical modeling was carried out using an artificial neural network – a very accurate tool which is popular nowadays”. This sentence is too concise and not precise, please change it.

The sentence corrected.

4. The caption under Fig. 4 is in bold. Figure 4. Change in the yield of rendered marmot fat with the Fan Boost enzyme preparation

Corrected

5. Dependence no. 1 has a wrong notation, if there is normalization, it must be an absolute value, otherwise it does not make mathematical sense

One of the types of normalization was used in the work: Min Max Scaling

6. All formulas are written in small font and are hard to read.

Font enlarged

7. In Fig. 10 I don't understand, at 70° there is a low concentration of enzymes, is this correct.If so, why.

The increase in fat yield using this enzyme preparation at a temperature above 70 0C is caused primarily by the temperature effect on the raw materials. During the experiment, no data were obtained on the yield of fat when using high concentrations of an enzyme preparation at a given temperature due to economic inexpediency

8. in the summary the authors write. “As a result of optimization, the data on the effect of temperature, process duration, introduced enzyme preparation and its concentration on the yield of rendered fat were obtained. there is nothing in the text about it.Please explain.

With the help of ANN, the optimal process parameters were obtained (paragraph 4.2). After that, we checked them in the experiment, did not write about it separately and indicated only in the conclusion.