search

vinaypedapati / AML-Survival

0 stars 0 forks source link

t-sne #2

Open vinaypedapati opened 1 year ago

StevenTheYi commented 1 year ago

Re-run t-sne with 6 month, 12 month, and 18 month survival

andysocial7 commented 1 year ago

Make sure you don't use regular overall survival but instead the binned survivals. Take the data points of each cluster and see what the commonalties are between those.

andysocial7 commented 1 year ago

Previous tsne was using total OS. If you wanted to make a model that takes all the gene mutations that predicts if someone is in cluster 1 or cluster 2.

Use survival data after you make the clusters. What I would want to do is do the 3 separate t-sne at 6, 12 and 18 mo intervals. Binned data has the same order as the mutations in the og aml prognosis file. K nearest neighbors is a separate clustering algorithm that woudn't require multiple T-SNEs

If we can get a nonobvious correlation in the cluster then that would be v valuable. Unclear why we think the binned data will have clearer clusters than the generalized data