Closed ccshao closed 1 year ago
Hi, the LIGER team have been actively incorporate new functions and features into the package. You are right about the original LIGER paper. The quantile normalization + Louvain method strategy was introduced afterwards along with online iNMF.
Thanks for the clarification, sorry I didn't notice the updates on the online algorithm.
The new algorithm is thoroughly described in the method section of our Nature Biotechnology paper (Gao et al 2021).
From: Chao Gao @.> Sent: Thursday, January 5, 2023 11:35:08 PM To: welch-lab/liger @.> Cc: Subscribed @.***> Subject: Re: [welch-lab/liger] "shared factor neighborhood" not used anymore as described in the paper? (Issue #280)
External Email - Use Caution
Hi, the LIGER team have been actively incorporate new functions and features into the package. You are right about the original LIGER paper. The quantile normalization + Louvain method strategy was introduced afterwards along with online iNMF.
— Reply to this email directly, view it on GitHubhttps://github.com/welch-lab/liger/issues/280#issuecomment-1373134411, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AAS2HVIEXHYN4FLVAVY42ATWQ6OHZANCNFSM6AAAAAATBNOV3E. You are receiving this because you are subscribed to this thread.Message ID: @.***>
Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues
Hi liger team.
Many thanks for this nice toolsets!
When I went through the methods parts of liger paper, the joint clustering is performed on the basis of shared factor neighborhood, e.g., get the factor neighborhood and calculate the Manhattan distance. However, it seems not to be the case anymore in the current implementation. Instead, it looks to me that Louvain community detection is performed on graphs built with knn on quantile normalized H matrix.
Did I misunderstood something? Details on the changes and implementations are appreciated, thanks a lot!