Alpha-1 Antitrypsin Deficiency

[rvida034]

The following data is focused on different zinc transporter genes and some metallothionein genes. There are different GSE series under investigation in which the wild type and PiZ mutation are compared via the following calculation: "t-Test: Paired Two Sample for Means".

[cwyoo]

The following data is focused on different zinc transporter genes and some metallothionein genes. There are different GSE series under investigation in which the wild type and PiZ mutation are compared via the following calculation: "t-Test: Paired Two Sample for Means".

Thank you for the update. Could you also post the information about the original dataset here for reference? Additionally, please share a message here outlining your plans for proceeding with the t-test results.

[rvida034]

Attached please find the document outlining the characteristics for the GEO DataSets that were analyzed. This document will be periodically updated.
Furthermore, the plan with the "t-Test: Paired Two Sample for Means" is to first determine which genes showed statistical significance (p-value < 0.05) (please refer to the "Research Project Data Updated" documented) to then narrow down which genes will be analyzed in the lab.

[rvida034]

Hi Samantha:

Could you please provide the Bayesian Network for the GEO DataSet series GSE141593 for only the 6-week old mice so that I may interpret it?

The genes of focus are as follows: SLC30A1, SLC30A2, SLC30A3, SLC30A4, SLC30A5, SLC30A6, SLC30A7, SLC30A8, SLC30A9, SLC39A1, SLC39A2, SLC39A3, SLC39A4, SLC39A5, SLC39A6, SLC39A7, SLC39A8, SLC39A9, SLC39A10, SLC39A11, SLC39A12, SLC39A13, SLC39A14, MTF1, MT1, MT2, MT3, and MT4.

Please let me know if you have any questions. Thank you so much!

[cwyoo]

Hi Samantha:

Could you please provide the Bayesian Network for the GEO DataSet series GSE141593 for only the 6-week old mice so that I may interpret it?

The genes of focus are as follows: SLC30A1, SLC30A2, SLC30A3, SLC30A4, SLC30A5, SLC30A6, SLC30A7, SLC30A8, SLC30A9, SLC39A1, SLC39A2, SLC39A3, SLC39A4, SLC39A5, SLC39A6, SLC39A7, SLC39A8, SLC39A9, SLC39A10, SLC39A11, SLC39A12, SLC39A13, SLC39A14, MTF1, MT1, MT2, MT3, and MT4.

Please let me know if you have any questions. Thank you so much!

Rebecca, I understand that Samantha has provided you with instructions for RNAseq data and Bayesian Networks analyses using R. Samantha will share these instructions in the "Manuscripts & Documentation" section, specifically under "Implemented Classes/Modules/Programs Documentation" titled "How to run scripts/programs." There, please click "Bayesian Network Analysis using R package bnlearn". Rebecca, please review the instructions and post there any challenges you encounter while following the steps. If you want to learn more about bnlearn, there is a separate forum created at Board index ‹ Manuscripts & Documentation ‹ Useful Tools for Statistical Analyses ‹ R package for Bayesian network structure learning (BNlearn).

If you prefer to use tools other than R, there are alternative tools available, e.g., Galaxy, Genie, Banjo, etc.. The goal is for you to comprehend the process so you can document and publish the results. Samantha, for the time being, could you conduct the analyses and share the results here?

Best,

[samanthagonzales]

Here is the full project folder for the initial Bayesian analysis on GSE141593, and the resulting network.
Please let me know if you have questions!

[cwyoo]

Here is the full project folder for the initial Bayesian analysis on GSE141593, and the resulting network.
Please let me know if you have questions!

Appreciate it, Samantha. Rebecca, could you please document here the data analysis process leading to the result? Typically, after cleaning RNAseq data, we start by creating a dataset containing genes with significant fold changes and genes of interest. Subsequently, we discretize the dataset using z-scores and identify the best-fitting Bayesian Networks.

For now, Dr. Luizzi and I have some comments and questions:
1. Kindly utilize gene names for the final network.
2. Are the genes showing significant fold changes included in this analysis?
3. Could you provide more details on the approach employed to identify the best-fitting Bayesian Networks? For instance, could you specify the duration and the number of runs conducted? Additionally, what criteria guided the decision to conclude the search?

[samanthagonzales]

Here is the full project folder for the initial Bayesian analysis on GSE141593, and the resulting network.
Please let me know if you have questions!

Appreciate it, Samantha. Rebecca, could you please document here the data analysis process leading to the result? Typically, after cleaning RNAseq data, we start by creating a dataset containing genes with significant fold changes and genes of interest. Subsequently, we discretize the dataset using z-scores and identify the best-fitting Bayesian Networks.

For now, Dr. Luizzi and I have some comments and questions:
1. Kindly utilize gene names for the final network.
2. Are the genes showing significant fold changes included in this analysis?
3. Could you provide more details on the approach employed to identify the best-fitting Bayesian Networks? For instance, could you specify the duration and the number of runs conducted? Additionally, what criteria guided the decision to conclude the search?

1. Kindly utilize gene names for the final network.

Can do!

2. Are the genes showing significant fold changes included in this analysis?

Only the genes of interest are included in this network, but I did write in the script the option to include top N differentially expressed genes. I can update the network to include those genes and post it here.

3. Could you provide more details on the approach employed to identify the best-fitting Bayesian Networks? For instance, could you specify the duration and the number of runs conducted? Additionally, what criteria guided the decision to conclude the search?

I tried different algorithms available to bnlearn, and hill-climbing was the only one to give me a completely directed network. I used the command `pzm.hc <- hc(pizmice.counts, restart = 500)` to generate the network.

hc() is the hill-climbing function provided by bnlearn. The default scoring is BIC, but it can be changed to BDE if that is preferred. Would it be better to run a for loop to create several structures, then choose the one with the best score (possibly with varying restart values)?

Let me know if you have further questions! I will include differentially expressed genes in the network and update the gene names in the meantime. Thanks!

[cwyoo]

Here is the full project folder for the initial Bayesian analysis on GSE141593, and the resulting network.
Please let me know if you have questions!

Appreciate it, Samantha. Rebecca, could you please document here the data analysis process leading to the result? Typically, after cleaning RNAseq data, we start by creating a dataset containing genes with significant fold changes and genes of interest. Subsequently, we discretize the dataset using z-scores and identify the best-fitting Bayesian Networks.

For now, Dr. Luizzi and I have some comments and questions:
1. Kindly utilize gene names for the final network.
2. Are the genes showing significant fold changes included in this analysis?
3. Could you provide more details on the approach employed to identify the best-fitting Bayesian Networks? For instance, could you specify the duration and the number of runs conducted? Additionally, what criteria guided the decision to conclude the search?

1. Kindly utilize gene names for the final network.

Can do!

2. Are the genes showing significant fold changes included in this analysis?

Only the genes of interest are included in this network, but I did write in the script the option to include top N differentially expressed genes. I can update the network to include those genes and post it here.

3. Could you provide more details on the approach employed to identify the best-fitting Bayesian Networks? For instance, could you specify the duration and the number of runs conducted? Additionally, what criteria guided the decision to conclude the search?

I tried different algorithms available to bnlearn, and hill-climbing was the only one to give me a completely directed network. I used the command `pzm.hc <- hc(pizmice.counts, restart = 500)` to generate the network.

hc() is the hill-climbing function provided by bnlearn. The default scoring is BIC, but it can be changed to BDE if that is preferred. Would it be better to run a for loop to create several structures, then choose the one with the best score (possibly with varying restart values)?

Let me know if you have further questions! I will include differentially expressed genes in the network and update the gene names in the meantime. Thanks!

Samantha and Rebecca, please read Learning Bayesian Networks from Data, A Practical Guide that I posted in this forum. Can you perform the search based on the posting and report the results? You may set max_hours to 32 and adjust the pecentage_difference based on the results you get between 1 and 2 hours. The document is work in progress, so please feel free to post any comments and questions there

[cwyoo]

Could you share approximately 200 genes that exhibit the most significant fold change? Please specify the threshold used to identify these significant genes. Additionally, please provide the thresholds for obtaining around 400 and around 100 significant genes.