Vorolign set

description of the benchmark set and the set itself

Big benchmark set

The benchmark set incorporates the consistently defined and classified domains in CATH 3.1.0 and SCOP 1.73. The domains used are filtered so that the maximal sequence identity between any domain pair is below 50%. The benchmark set contains domains from 2043 SCOP families , 1139 SCOP superfamilies and 672 SCOP folds.

• similar pairs (same SCOP fold and same CATH topology): 129436 Domain pairs
• non-similar pairs (different folds in SCOP and different topologies in CATH): 1740476 Domain pairs

Download: clusters of domains with minimum 50% sequence identity.
The file determines for each representative domain (used in the benchmark set) the domains that share at least 50% sequence identity with the cluster representant (if any) which may be used to use different or additional domains for benchmarking purposes.

The full mapping between CATH and SCOP.

Vorolign set

Big set

To map the a tuple (length, RMSD) to a single similarity value, we use two functions to assign lower and upper RMSD bounds for every possible block length. A block with length l below the lower bound will get the similarity value 1.0, blocks with higher RMSD values than the upper bound are rejected (i.e. similarity value 0.0).
The same similarity function applies for the block pair topology scoring, where the RMSD is computed as the maximal RMSD of the the two blocks in the implied superposition.

The cost of a topological mutation is cost=4*(1.0-sim(MAX(rmsd1, rmsd2), length)).

The upper bound is defined by the function:


The lower bound is defined by the function:


The lower and upper bounds are determined by the functions shown on the figure below.

The similarity value for a block having RMSD r between the lower and upper bound is simply proportional to the difference of (r-lower)/(upper-lower). The resulting similarity functions for different lengths l are shown below.