SMLG (Statistical Machine Learning Group) Discussion Forum

Here is the gene list, excel spreadsheet, and R Data object for varying subsets of differentially expressed genes (100,200,400)
First, significant differential expression results were extracted from DESeq2 by using an adjusted p-value threshold of .05
From the subset of genes whose p_adj < .05, I used varying log2FC thresholds to obtain N differentially expressed genes

The absolute log2FC thresholds to obtain N genes are:
N = 100, abs(log2FC) > 2.58 --
N = 200, abs(log2FC) > 1.911 --
N = 400, abs(log2FC) > 1.4555 --

I then took the union of the N differentially expressed genes with the 17 genes of interest present in GSE141593 to obtain the dataset for the network analyses, for a max of N + 17 genes per dataset.

Let me know if you have any questions. Thanks!

samanthagonzales wrote:Here is the gene list, excel spreadsheet, and R Data object for varying subsets of differentially expressed genes (100,200,400)
First, significant differential expression results were extracted from DESeq2 by using an adjusted p-value threshold of .05
From the subset of genes whose p_adj < .05, I used varying log2FC thresholds to obtain N differentially expressed genes

The absolute log2FC thresholds to obtain N genes are:
N = 100, abs(log2FC) > 2.58 --

GSE141593_log2fc_thresh=2.58_ngene=117_zn_transporter_gene_list.txt

N = 200, abs(log2FC) > 1.911 --

GSE141593_log2fc_thresh=1.911_ngene=217_zn_transporter_gene_list.txt

N = 400, abs(log2FC) > 1.4555 --

GSE141593_log2fc_thresh=1.4555_ngene=416_zn_transporter_gene_list.txt

I then took the union of the N differentially expressed genes with the 17 genes of interest present in GSE141593 to obtain the dataset for the network analyses, for a max of N + 17 genes per dataset.

Let me know if you have any questions. Thanks!

Please note that the N=200 dataset is different than the dataset used in the first analyses, as for the original dataset I sorted the differentially expressed genes according to increasing p-value and took the first 200. I will compare the overlap of genes between these two sets and post them here.

samanthagonzales wrote:

samanthagonzales wrote:Here is the gene list, excel spreadsheet, and R Data object for varying subsets of differentially expressed genes (100,200,400)
First, significant differential expression results were extracted from DESeq2 by using an adjusted p-value threshold of .05
From the subset of genes whose p_adj < .05, I used varying log2FC thresholds to obtain N differentially expressed genes

The absolute log2FC thresholds to obtain N genes are:
N = 100, abs(log2FC) > 2.58 --

GSE141593_log2fc_thresh=2.58_ngene=117_zn_transporter_gene_list.txt

N = 200, abs(log2FC) > 1.911 --

GSE141593_log2fc_thresh=1.911_ngene=217_zn_transporter_gene_list.txt

N = 400, abs(log2FC) > 1.4555 --

GSE141593_log2fc_thresh=1.4555_ngene=416_zn_transporter_gene_list.txt

I then took the union of the N differentially expressed genes with the 17 genes of interest present in GSE141593 to obtain the dataset for the network analyses, for a max of N + 17 genes per dataset.

Let me know if you have any questions. Thanks!

Please note that the N=200 dataset is different than the dataset used in the first analyses, as for the original dataset I sorted the differentially expressed genes according to increasing p-value and took the first 200. I will compare the overlap of genes between these two sets and post them here.

Thanks, I suggest we use abs(log2FC) > 2.0 as a threshold.

Sounds good! Attached is the updated data using abs(log2fc) > 2.0. (N = 197 genes)

I also compared the original dataset generated by taking the top 200 genes with the smallest p-value versus this new dataset.
There are 115 genes in common, 102 in the top 200 subset, and 82 unique to the log2fc subset. I've attached a Venn Diagram illustrating the comparison as well as a spreadsheet indicating which genes are in which set.



Let me know if you have any questions!

Raw Banjo analysis results (32h) on newest data (log2fc < 2.0) 197 genes, 10 observations

I have attached the color code (function for the genes involved in the top enriched processes GO)

jliuzzi wrote:I have attached the color code (function for the genes involved in the top enriched processes GO)

Below is a list of networks derived from the banjo results on GSE141593:
1. The complete network, including arcs weighted by influence score (positive scores are blue, negative scores are orange-dashed, 0 scores are black-dotted), and genes color coded by their pathway annotations that Dr. Liuzzi provided.
2. The markov blanket of "pizMutation", using bnlearn's mb() function -- color coded by the provided GO annotations
3. The secondary markov blanket subnetwork, using bnlearn's mb() function on each node in the primary markov blanket, color coded by the provided GO annotations.

For genes with multiple pathway annotations, I took the first pathway for simplicity. I will have to spend a little more time to find a way to include all the annotations in a coherent way for future iterations of the graphs.
Your feedback is greatly appreciated!
Thanks

Here are the primary and secondary Markov blankets annotated by their influence scores, extracted from the full network.

Could it be possible to post a subnetwork with the second makrov blanket and add this genes: SLC39a11, Slc39A14, TFEB, TFE3, Becn1, lamp1, lamp2, Hspa8, Atg5, Atg7, Atg12, Atg16L, Map1lc3b, Traf2, and Nfkb?

Thank you

jliuzzi wrote:Could it be possible to post a subnetwork with the second makrov blanket and add this genes: SLC39a11, Slc39A14, TFEB, TFE3, Becn1, lamp1, lamp2, Hspa8, Atg5, Atg7, Atg12, Atg16L, Map1lc3b, Traf2, and Nfkb?

Thank you

Here is the second degree Markov blanket with the requested genes included.

A majority of the genes were not in the original network of 198 genes, so I reran banjo to include them for a new total N = 211. The raw banjo results for that run are also attached.
2 genes as listed (Nfkb, Atg16L) could not initially be found in the data, but after checking the Mouse Genome Informatics database(https://www.informatics.jax.org/) and NCBI Gene (https://www.ncbi.nlm.nih.gov/gene?cmd=Retrieve), I found their aliases/mouse orthologs: Nfkb1, Atg16l1

Please let me know if you have any questions via Teams - thanks!