Skip to main content
. 2014 Sep 6;11(18):47–58. doi: 10.1016/j.csbj.2014.09.001

Table 1.

Assessing the effectiveness of 3 distance measures (FASTA, STRUCTAL on 3D sequence traces, and STRUCTAL on 3D structures) in identifying proteins with similar structures.


MDS spacea

FASTA
3DSEQ
3DSTRUCT
Cluster 2 3 4 5 2 3 4 5 2 3 4 5
1 30.8b 36.6 43.6 47.6 34.5 41.7 47.0 50.6 39.7 60.7 62.2 77.0
2 32.6 38.9 45.2 49.5 31.3 39.7 46.0 50.2 56.8 71.0 72.2 76.6
3 27.7 34.4 41.4 45.5 44.1 50.9 56.7 59.4 82.5 87.1 88.00 90.9
4 28.3 34.9 42.4 47.3 31.8 40.0 45.5 49.3 40.7 54.5 56.9 71.2
5 28.7 35.0 41.3 45.6 33.2 42.2 49.3 52.9 50.2 67.0 70.4 74.6
AIS 29.6c 35.9 42.8 47.1 35.0 42.9 48.9 52.5 54.0 68.0 70.0 78.1
a

Dimension P of the MDS projected space, i.e. number of coordinates representing each protein. Fig. 7 provides illustrations of this space for P = 2.

b

Average distance dav = 0.25 ∑ j ≠ id(CiCi), where d is computed using Eq. (10). The distances are scaled as fraction of the maximum distance between any two proteins in CATH605.

c

Average Inter-cluster Separation (AIS) computed using Eq. (10). A large value for AIS indicates a good separation between the clusters.