Protein database searches using compositionally adjusted substitution matrices

Kewl you should come up with that. Execllent!

[edit] Compositional distance

Distance metric for two probability distributions:

Failed to parse (Cannot write to or create math temp directory): D^2(\vec{r},\vec{s}) = \frac{1}{2} \sum_{i} (r_{i} ln \frac{2 r_i}{r_i + s_i} + s_{i} ln \frac{2 s_i}{r_i + s_i})

• Compositional adjustment is on average preferred for two sequences if their compositions Failed to parse (Cannot write to or create math temp directory): \vec{r}

and Failed to parse (Cannot write to or create math temp directory): \vec{s}
have a distance D less than 0.16

• NOTE: they cite a paper for this metric.