Fast structure similarity searches among protein models: efficient clustering of protein fragments

Table 4 Fragment clustering on backbone.

RMSD threshold (Å)	N. struct.	This work	$\frac{N (N - 1)}{2}$	ratio	Rep. Struct.
0.05	107,184	53,793,796	5,744,151,336	0.0094	105,299
0.1	105,299	124,506,221	5,543,887,051	0.0225	87,195
0.2	87,195	57,001,567	3,801,440,415	0.0150	63,829
0.4	63,829	65,723,340	2,037,038,706	0.0322	21,637
0.8	21,637	33,375,313	234,069,066	0.1426	2,445
1.6	2,445	2,586,436	2,987,790	0.8657	33
3.2	33	528	528	1.0000	2
Total	107,184	336,987,202	5,744,151,336	0.0587

Columns report the threshold RMSD chosen for clustering, the number of starting fragments, the number of RMSD computations done, the number of computations in an all-vs-all comparison, the number of representative fragments, used as starting fragments at the next iteration. Atoms N, CA, C, O have been used for superposition.

ISSN: 1748-7188