search

JianiC / RSV_Epitope

5 stars 2 forks source link

MDS vs. UMAP #7

Open JianiC opened 3 years ago

JianiC commented 3 years ago

MDS vs. UMAP UMAP in R: https://cran.r-project.org/web/packages/umap/vignettes/umap.html Input: Smith et al; epitope comparison with vaccine cluster: k-means n= 11 empirical seed =123

Cluster determine by MDS smith_vaccine_mds validate in UMAP ump

Discrepancies observed!

_Originally posted by @JianiC in https://github.com/JianiC/RSV_Epitope/issues/5#issuecomment-766541862_

JianiC commented 3 years ago

Using MDS to study T-cell epitope for influenza has been published, this paper also studies the change between subtype https://www.nature.com/articles/srep33334?proof=t @swantan

  1. From this paper, I think we should not focus on k-means clustering, we actually only use it for visualization, and almost everyone uses empirical information. Maybe we should focus on the movement in the time scale, maybe get the centroid for the sequence in the same year.
  2. I am also considering removing the phylogeny mapped epitope part in my RSV work, maybe from immuno-scale jump to conserved epitope directly. 9-mer peptides seem to make things more complicated, but we do need to identify human cross epitope and evaluate conservancy in JanuxMatrix
  3. For T-cell immunity estimation, instead of measuring the distance from the detention reduction visualization, we should go to the raw data just get the proportion of non-cross T-cell immunity, what about " loss of cross T-cell immunity" instead of T-cell immunity distance?
JBahl commented 3 years ago

Nice to see we have something to cite.

What do you mean by ‘movement in the timescale'?

On Jan 25, 2021, at 10:41 PM, JianiC notifications@github.com<mailto:notifications@github.com> wrote:

[EXTERNAL SENDER - PROCEED CAUTIOUSLY]

Using MDS to study T-cell epitope for influenza has been published, this paper also studies the change between subtype https://www.nature.com/articles/srep33334?proof=t @swantanhttps://github.com/swantan

  1. From this paper, I think we should not focus on k-means clustering, we actually only use it for visualization, and almost everyone uses empirical information. Maybe we should focus on the movement in the time scale, maybe get the centroid for the sequence in the same year.
  2. I am also considering removing the phylogeny mapped epitope part, maybe from immuno-scale jump to conserved epitope directly. 9-mer peptides seem to make things more complicated, but we do need to identify human cross epitope and evaluate conservancy in JanuxMatrix

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/JianiC/RSV_Epitope/issues/7#issuecomment-767275805, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ADSPRN2A37KHUXNHTK3UQPLS3Y2WNANCNFSM4WRFI4WQ.