I’m struggling to interpret the DE results I get with scVI model. I’m looking for changes in expression between two conditions in a cluster. All bayes_factor values in the result are < 3. I looked for guidelines and found the DE in worm tutorial. There’s a table there which is supposed to help with filtering results, I suppose. The tutorial stresses that the returned bayes_factor for each gene is ln(BF_10). Now, according to the table, bayes_factor > 1.1 would be “Moderate evidence for H1”. This would be a reasonable cut-off I’d be willing to use. However, I’d love to know more about your real-world experience. How do you filter your DE results?
Best