Open yintz opened 1 year ago
@GeorgescuC any idea about this? is there any better cluster methods can cluster like the above? thank you
Hi @yintz ,
I see in the first figure that you are only looking at 1 chromosome (and one of the smaller ones). Was the clustering run on the data that includes only that chromosome, or was it run on the data of all chromosomes? There are options you can experiment with to try to improve the clustering since the best ones can vary between datasets, specifically the "hclust_method" which as the name implies controls the parameter of hclust(), and accepts the values "ward.D", "ward.D2", "single", "complete", "average", "mcquitty", "median", "centroid". If you are using one of the subclustering methods to pre-split the cells, those also have options to control the granularity, with either "tumor_subcluster_pval" or "leiden_resolution".
Regards, Christophe
Dear,
I just checked the clone cluster effect on chromosome level. it look like below. I also tried different cluster methods mentioned in the function. but none of them could cluster all the similar clone together. (like the the same length of blue or red together. ) if they can not cluster all the similar pattern together in one chromosome level, how could we believe it could cluster all the similar patten for all chromosome?
this is not well clustered in one chromosome level
do you think the ideal cluster should be like this? cluster all the same length clone together?