@isaacovercast @dereneaton I found that the command -m was very useful! Now, I separate my data to several pools and run step 1-4 respectively. Subsequently, I want to merge all the assemblies to a new one and then run step 5-7. Also, I detected the different flows with the same small database and got the same results. But I want to know if I run a very big dataset in this way, it will get the same result comparing with pooling the data together?