examl1

Exascale maximum likelihood (examl) code for phylogenetic inference examl-avx - exascale maximum likelihood (examl) code for phylogenetic inference using avx examl-omp - exascale maximum likelihood (examl) code for phylogenetic inference using omp examl-omp-avx - exascale maximum likelihood (examl) code for phylogenetic inference using avx and omp

Usage

examl { | examl-AVX | examl-OMP | examl-OMP-AVX -s binarySequenceFileName -n outputFileNames -m rateHeterogeneityModel -t userStartingTree | -R binaryCheckpointFile | -g constraintTree -p randomNumberSeed [-np NP] [-a] [-B numberOfMLtreesToSave] [-c numberOfCategories] [-D] [-e likelihoodEpsilon] [{-f d | e | E | o | q}] [-h] [-i initialRearrangementSetting] [-I quartetCheckpointInterval] [-M] [-r randomQuartetNumber] [-S] [-v] [-w outputDirectory] [-Y quartetGroupingFileName] [{--auto-prot=ml | bic | aic | aicc}]}

Options

-np NP: NP is the number of processors to use for MPI. If not specified some hopefully sensible guess is made. This is a specific option of the Debian wrapper in /usr/bin.

-a: use the median for the discrete approximation of the GAMMA model of rate heterogeneity

: DEFAULT: OFF

-B: specify the number of best ML trees to save and print to file

-c: Specify number of distinct rate catgories for ExaML when modelOfEvolution is set to GTRPSR Individual per-site rates are categorized into numberOfCategories rate categories to accelerate computations.

: DEFAULT: 25

-D: ML search convergence criterion. This will break off ML searches if the relative Robinson-Foulds distance between the trees obtained from two consecutive lazy SPR cycles is smaller or equal to 1%. Usage recommended for very large datasets in terms of taxa. On trees with more than 500 taxa this will yield execution time improvements of approximately 50% While yielding only slightly worse trees.

: DEFAULT: OFF

-e: set model optimization precision in log likelihood units for final optimization of model parameters

: DEFAULT: 0.1

-f: select algorithm:

"-f d": new rapid hill-climbing: DEFAULT: ON

: "-f e": compute the likelihood of a bunch of trees passed via -t

: this option will do a quick and dirty optimization without re-optimizng the model parameters for each tree

: "-f E": compute the likelihood of a bunch of trees passed via -t

: this option will do a thorough optimization that re-optimizes the model parameters for each tree

: "-f o": old and slower rapid hill-climbing without heuristic cutoff

: "-f q": fast quartet calculator

: DEFAULT for "-f": new rapid hill climbing

-g: Pass a multi-furcating constraint tree to ExaML. The tree needs to contain all taxa of the alignment! When using this option you also need to specify a random number seed via "-p"

-h: Display this help message.

-i: Initial rearrangement setting for the subsequent application of topological changes phase

-I: Set after how many quartet evaluations a new checkpoint will be printed.

: DEFAULT: 1000

-m: Model of rate heterogeneity

: select "-m PSR" for the per-site rate category model (this used to be called CAT in RAxML) select "-m GAMMA" for the gamma model of rate heterogeneity with 4 discrete rates

-M: Switch on estimation of individual per-partition branch lengths. Only has effect when used in combination with "-q" Branch lengths for individual partitions will be printed to separate files A weighted average of the branch lengths is computed by using the respective partition lengths

: DEFAULT: OFF

-n: Specifies the name of the output file.

-p: Specify a random number seed, required in conjunction with the "-g" option for constraint trees

-R: read in a binary checkpoint file called ExaML_binaryCheckpoint.RUN_ID_number

-r: Pass the number of quartets to randomly sub-sample from the possible number of quartets for the given taxon set. Only works in combination with -f q !

-s: Specify the name of the BINARY alignment data file generated by the parser component

-S: turn on memory saving option for gappy multi-gene alignments. For large and gappy datasets specify -S to save memory This will produce slightly different likelihood values, may be a bit slower but can reduce memory consumption from 70GB to 19GB on very large and gappy datasets

-t: Specify a user starting tree file name in Newick format

-v: Display version information

-w: FULL (!) path to the directory into which ExaML shall write its output files

: DEFAULT: current directory

-Y: Pass a quartet grouping file name defining four groups from which to draw quartets The file input format must contain 4 groups in the following form: (Chicken, Human, Loach), (Cow, Carp), (Mouse, Rat, Seal), (Whale, Frog); Only works in combination with -f q !

--auto-prot=ml|bic|aic|aicc When using automatic protein model selection you can chose the criterion for selecting these models.

: RAxML will test all available prot subst. models except for LG4M, LG4X and GTR-based models, with and without empirical base frequencies. You can chose between ML score based selection and the BIC, AIC, and AICc criteria.

: DEFAULT: ml