Introduction
The Quasar system was designed to make it possible to explore, benchmark, optimize and use various methods to rank comparative modelling and fold recognition alignments (also called sequence-structure alignments) easily with respect to a user defined scoring formula.
In order to support the process of finding and using such a formula, Quasar offers the possibility to combine one or more scoring schemes in a linear or non-linear formula with help of a score conductor that can be used afterwards to calculate a quality score for all alignments in an alignment file.
Quasar already provides a number of common alignment quality measurements (= scoring schemes) like different amino acid exchange matrices (PAM, Blosum, conformational matrices...) and secondary structure measures like SOV or Q3.
Scoring schemes can easily be combined in a scoring formula that is represented by a score conductor, e.g. a weighted sum, in the Quasar environment. Additionally Quasar provides the possibility to add new scoring schemes or score conductors to the system as well as to the graphical user interface (GUI) such that Quasar can easily be extended to fit your needs.
Another feature of Quasar is that a number of benchmark scores like TOUCH, APDB, MaxSub... is also available in the system. That way you can easily test the preformance your new scoring scheme or scoring formula on a benchmark set (where query and template structure are known) compared to real quality measurements like RMSD, contact matrix similarity....
To optimize the weights assigned to each individual scoring scheme in a score conductor, Quasar contains two optimization algorithms, namely a Genetic Algorithm and a Least Squares Optimization. This feature can help users to even further improve the ranking performance of a scoring function.