Abstract
Transmembrane proteins are estimated to represent more than a quarter of\r the human proteome and make up half of all pharmaceutical drug targets. However, \r it is significantly more difficult to study membrane proteins than soluble\r proteins. For instance, while the rate of sequencing of membrane proteins\r is similar to that of soluble proteins they make up less than 1% of solved\r protein structures. Evolutionary models are often used to group biological\r sequences together in order to facilitate the development of hypotheses.\r This is especially useful for transmembrane proteins in order to correlate\r sequences with known structures to sequences without solved structure. We\r have developed a set of alpha-helical membrane protein specific replacement\r matrices used for the a priori fixing of evolutionary parameter space for\r these protein types. The implementation of these partitioned matrices in phylogenetic\r software more accurately infers phylogenies for transmembrane protein sequence\r alignments. The matrices extract more data from the sequence information\r in comparison to matrices considered to be standards of the field. Structural\r information was used to inform the calculation of these matrices and we show\r that doing so significantly increases the amount of evolutionary information\r in the values of the replacement matrix. Our matrices and inference methods\r increase the biological accuracy of phylogenies calculated for alpha helical\r membrane proteins without an increase in the number of parameter.