MrBayes blocks
http://darwin.zoology.gla.ac.uk/~rpage/mrbayes

		Sample block

begin mrbayes;
	[The following line is useful for automatic execution with no
	 warnings issued before files are overwritten and automatic
	 termination of chains after the prespecified number of generations.]
	set autoclose=yes nowarn=yes;
	
	[The following block demonstrates how you can set up a model that
	 allows the third codon position to have gamma-distributed rate
	 variation with a different shape parameter than the first and 
	 second positions.]
	charset first_pos  = 1-720\3;
	charset second_pos = 2-720\3;
	charset third_pos  = 3-720\3;
	partition by_codon = 3:first_pos,second_pos,third_pos;
	

	[Sets starting tree. If no tree is defined, a random tree is used.]
	usertree = (((FR:0.121947,MS2:0.085573):0.164488,GA:0.245551):0.215261,
	           ((SP:0.108856,NL95:0.111132):0.106193,((M11:0.061203,MX1:0.100377):
	           0.113943,QB:0.131502):0.091978):0.281641,PP7:0.465396);	
	
	["sitepartition=bycodon" sets the name of the site partition to use for site specific rate variation as "bycodon"]

	set partition=by_codon;
	
	[sets the parameters of the likelihood model."nst=6" sets the general time reversible model.]

	lset nst=6 rates=gamma; [ Suggested starting model]
	
	[alternatively, if you want to apply different models to different paritions of the data, for instance a different model for 1st & 2nd codon positions vs 3rd codon positions
		lset applyto=(1,2) nst=2	
		lset applyto=(3)   nst=6	os números entre parenteses correspondem à ordem na
						definição da partition "by_codon=3:..."

	you can also use applyto=(all), attempts to apply parameter settings to all of the data partitions.
 
	[This is the line that allows the gamma shape parameter of the third partition to be unique:]
	unlink shape=(3);

	[run for "ngen", print results to screen every "printfreq", save current tree to file every 	"samplefreq", run "nchains" simultaneously, save branch lengths]
	
	mcmc ngen=1000 printfreq=100 samplefreq=100 nchains=4 savebrlens=yes  burnin=00;

	[burnin=number of trees to be ignored. contype is the consesus type. Set to "halfcompat" 	returns the equivalent of 50% majority rule in PAUP*, "allcompat" returns equivalent to 50% 	majority rule with 'show frequencies of all observed biparitions' ticked.]

	sumt filename=replicase.nex.t contype=allcompat burnin=00;
			[sumt and sump have SEPARATE BURNIN times]


end;