Hello. My lab has recently begun using Kilosort, and I am doing a comparison between MClust and Kilosort. I cluster my files with both programs then run a correlation algorithm to match cells found in both programs- with the hopes that we find mostly the same cells no matter how we do it. Some files are spot on, and there is a large overall correlation between clusters in both programs. However, I have encountered two files where I can produce the same number of cells in both MClust and Kilosort (28 to be exact) yet when I run the correlation algorithm less than half of the cells match up. I have adjusted ops.Th, ops.lam, ops.mergeT, and ops.splitT all to varying degrees without seeing a significant difference in my clusters and thus no difference in my correlation algorithm. Do you have any advice on which parameters to adjust and how I could adjust them to produce cleaner results?
Hello. My lab has recently begun using Kilosort, and I am doing a comparison between MClust and Kilosort. I cluster my files with both programs then run a correlation algorithm to match cells found in both programs- with the hopes that we find mostly the same cells no matter how we do it. Some files are spot on, and there is a large overall correlation between clusters in both programs. However, I have encountered two files where I can produce the same number of cells in both MClust and Kilosort (28 to be exact) yet when I run the correlation algorithm less than half of the cells match up. I have adjusted ops.Th, ops.lam, ops.mergeT, and ops.splitT all to varying degrees without seeing a significant difference in my clusters and thus no difference in my correlation algorithm. Do you have any advice on which parameters to adjust and how I could adjust them to produce cleaner results?