![]() |
![]() |
|
Here we attempt to evaluate individual methods by combining several characteristics together. This gives a more complex and comprehensive look at the performance of methods.
Values of each of the characteristics contributing to the composite evaluation. The three features above are weighted and combine to produce the final score, which is a composite evaluation score. ![]() Composite evaluation of methods. Three characteristics are weighted as follows: fraction of correctly assigned chains: 60%, fragmentation of discontinuous domains: 20%, precision of domain overlap:20%. The overall performance is represented in percent. We varied the weighting schema and found that regardless of the placed weights, PDP method appears to be superior and PUU method is inferior. However, DomainParser and NCBI methods trade places depending on whether fragmentation and precision of domain overlap (strong points of DomainParser) or correct assignment of domains (strong point of NCBI method) are weighted more heavily. ![]()
Composite evaluation of methods under different weighting schemas. The weighting parameters are as follows: first parameter: : fraction of correctly assigned chains, second parameter: fragmentation of discontinuous domains, third parameter: precision of domain overlap. |
This work is sponsored by the National Institutes of Heath (NIH) Grant Number GM63208 (NIH/NIGMS)