Pages

Tuesday, August 7, 2012

Study of Uclust Vs. DA-PWC for Divergent Dataset

Description DA-PWC Uclust 0.65 Uclust 0.75 Uclust 0.85 Uclust 0.95
Total number of clusters 23 4 10 36 91
Total number of clusters uniquely identified i.e. one original cluster goes to one uclust cluster 23 0 0 13 16
Total number of shared clusters with significant sharing (one uclust cluster goes to > 1 real cluster 0 4 10 5 0
Total number of uclust clusters that are just part of a real cluster 0 4 10 17(11) 72(62)
Total number of real clusters that are one uclust cluster but uclust cluster is spread over multiple real clusters 0 14 9 5 0
Total number of real clusters that have significant contribution from > 1 uclust cluster 0 9 14 5 7


Associated Histograms

Note. Uclust and CD-HIT were run with a similarity cut of 0.97 (default), which denotes that two sequences belong to different clusters if they have a similarity value less than 0.97.


  • Divergent Dataset


  • Artificial Dataset

Analysis of Divergent Dataset


Description

Here 23 True Sequences are associated with 23 clusters found accurately by DAPWC Deterministic Annealing Pairwise Clustering

Links

Images

Presented below are the images of length dependence and clusters from DA-PWC and Uclust 0.85 and 0.95. Note statistics is not very good and so some refinements are not realistic.

  • Length Study




  • DA-PWC Clusters




  • Uclust 0.85


Analysis of Artificial Dataset


Description

Here there are 89 True Sequences and around 30 identifiable clusters. As seen in plots some clusters have multiple True Sequences near them -- presumably separate families are not resolvable, a few True Sequences seem far from any sequences. All determined clusters (from DA-PWC) have one or more True Sequence reasonably near.


Links

Images

Presented below are the images of length dependence and clusters from DA-PWC and Uclust 0.85 and 0.95. Note statistics is not very good and so some refinements are not realistic.
  • Length Study




  • DA-PWC Clusters




  • Uclust 0.85




  • Uclust 0.95




Monday, April 9, 2012