Hi Sarah

instead of arbitrary reducing the list, could you only choose FC and adjusted Pval to refine the results to a gene list to be analyzed as a whole, without having an eye on the names and their specificities?

all the best

fred

Frederic Lepretre i have list of all genes with their p values and fdr . now i am confused how to choose top 200 if i want to write a publication.

I have also drawn the heat map for top 200 can i use these heat maps 200 genes for further gene ontology and pathway enrichment and then connect with my research topic. kindly give me some suggestions. thanks

Hi Sarah

for publication I wonder if the limitation of 200 genes is relevant to your project and results. I don't know since it depends also on the quality of your sequencing and results.

2000 unregulated genes is high since it is known that about 1000 to 5000 genes are expressed in cells.

can you consider a threshold of 2 for the FC and the lowest Pval and list the number of genes passing though the filters. if not sufficient, get higher filters.

apply different FC and adjPval and make a table. in regards of your data, qualities and hypothesis, it maybe help you to select the right list of genes to be analyzed.

last, do not forget the downregulated genes, they also their role in pathogenic biology development.

all the best

fred

Frederic Lepretre Thanks I will consider these tips for further selection of genes

I'd suggest to do GO term and network analyses first on all 2000 genes. Depending on results from these analyses, you can then focus on some key genes (such as hub genes from network analysis). Such analyses are more helpful with more genes than simply the top 100 genes. It might be more helpful to let the function/network results guide you instead of filtering on an individual gene basis.

Xinkun Wang thanks For your answer. It is really helpful. I have one confusion that after network analysis of my 2000 upregulated genes how can I select the hub genes. As I have read in most of the papers they select hub genes and then they discuss in the paper. Kindly guide me in the selection of hub genes.

Hi Sarah,

I agree with the above answers. You don't want to limit yourself to the top 100 or 200 genes just because this might the maximum number that visualize well. What was the p-value used in your DE analysis? You may want to filter on a slightly more rigorous adjusted p-value than the default in whatever differential expression software you used.

What was your hypothesis for this experiment regarding the DE analysis?

As mentioned above you need to analyze the upregulated and downregulated genes. The simple difference of which order you entered your two groups of samples into DE analysis software would swap these, so the biology of both should be considered.

I'm not an expert on network analysis, but I think you'll want to take the top ~10-20% of genes with the most connections as your hub genes. Searching the literature for cut-offs used by others would be helpful.

Cameron J Grisdale Thanks for the answer. The P-value selected for the analysis is 0.05. I am looking at the genes that are upregulated into osteoclast differentiation and then focus on the role of these genes in osteoporosis. I have a list of total 5000 DEG in which 2300 upregulated and rest are downregulated. Now I at network analysis and trying to find out the selection of hub genes.

2 fold up-regulated and 2 fold down-regulated genes are considered for GO and network analysis along with pvalue 0.05 and fdr value or q value 0.05

Hi Sarah,

As for heatmap visualization, you could select the top 100 or 200 DEGs to show the differential expression. I think it's a right and wise way. By the way, you could change the cut-off (fold change/logFC, p-value/q-value) arbitrarily.

However, if you would like to conduct functional analysis such as GO or KEGG, I recommand you should input all DEGs into software or webpage (DAVID or R). The reason is that most biologic variations derive from the transcriptome changes, which include signaling pathways cascade and downstream effectors.

So, if you want to further narrow down, network analysis is indeed a good way. Cytoscape software and its additional plugins will be a better choice for you to find the hub gene.