I'm interested in applying WGCNA to my dataset, but I'm unsure whether I could apply it, the dataset consists of 3 different clones with diverse response to treatment, from resistant to sensitive. Several treatment conditions are analyzed: Ig2b (control for combo), combo, control (control for actla4) , actla4. However, I don't have enough samples for performing one network for each condition, Would it be appropriate to combine all treatments and controls into a single dataset to improve the analysis, or could this introduce confounding effects?
My dataset is: "C11_IgG2b_E1", "C11_IgG2b_E2" , "C11_IgG2b_E3", "C11_aCTLA4_E4" , "C11_aCTLA4_E5" ,"C11_aCTLA4_E6", "C11_aCTLA4_E7" , "C11_aCTLA4_E8" , "C11_Ctr_E9" ,"C11_Ctr_E10" , "C11_Ctr_E11" , "C11_Combo_E12" , "C11_Combo_E13" , "C11_Combo_E14" , "C11_Combo_E15" , "C11_Combo_E16" , "C14_IgG2b_E17" ,"C14_IgG2b_E18" , "C14_IgG2b_E19" , "C14_aCTLA4_E20" ,"C14_aCTLA4_E21" ,"C14_aCTLA4_E22" ,"C14_aCTLA4_E23" ,"C14_aCTLA4_E24" ,"C14_Ctr_E25" ,"C14_Ctr_E26" , "C14_Ctr_E27" ,"C14_Combo_E28" , "C14_Combo_E29" , "C14_Combo_E30" , "C14_Combo_E31" , "C14_Combo_E32" , "C15_IgG2b_E33" "C15_IgG2b_E34" , "C15_IgG2b_E35" , "C15_aCTLA4_E36" ,"C15_aCTLA4_E37" ,"C15_aCTLA4_E38" ,"C15_aCTLA4_E39", "C15_aCTLA4_E40" ,"C15_Ctr_E41","C15_Ctr_E42","C15_Ctr_E43","C15_Combo_E44" , "C15_Combo_E45" , "C15_Combo_E46" , "C15_Combo_E47" , "C15_Combo_E48"
•
written
by
María José •
2 answers
I would probably build a single network using all samples rather than constructing separate networks for each condition, as the sample size per condition is quite limited. After module detection, you can correlate module eigengenes with treatment, clone, or resistance phenotypes to identify relevant biological signals. Unless your goal is module preservation analysis, combining all samples into one network is likely the most practical approach here.
Happy to discuss further. https://github.com/hongthai-lenguyen
A single network for all samples. Then a cluster analysis will probably yield sample/condition-specific clusters. I had a similar issue with sample size and for my analysis, I could get condition-specific clusters. I expanded the clusters by using those genes as "seed" to pool in protein-protein interactions, and further did some pathway analysis. SO then, ultimately, I did have individual networks of protein-protein interactions Hope this helps,
Sanchari
Log in to answer this question.