RAXMLHPC(1)

📦
version 8.2.13 · 2 versions
, in debian trixie, ubuntu questing, ubuntu resolute, ubuntu stonking

okay bestie, this is the manual page for raxmlHPC(1) — Randomized Axelerated Maximum Likelihood. It's giving phylogenetics main character energy, and we're going section by section. 💅✨

🌸 NAME 🌸

raxmlHPC — Randomized Axelerated Maximum Likelihood

🦋 SYNOPSIS 🦋

raxmlHPC
[-SSE3|-AVX|-PTHREADS|-PTHREADS-SSE3|-PTHREADS-AVX|-HYBRID|-HYBRID-SSE3|HYBRID-AVX]
-s sequenceFileName -n outputFileName -m substitutionModel [-a
weightFileName] [-A secondaryStructureSubstModel] [-b
bootstrapRandomNumberSeed] [-B wcCriterionThreshold] [-c
numberOfCategories] [-C] [-d] [-D] [-e likelihoodEpsilon] [-E
excludeFileName] [-f
a|A|b|B|c|C|d|D|e|E|F|g|G|h|H|i|I|j|J|k|m|n|N|o|p|P|q|r|R|s|S|t|T|u|v|V|w|W|x|y]
[-F] [-g groupingFileName] [-G placementThreshold] [-h] [-H] [-i
initialRearrangementSetting] [-I autoFC|autoMR|autoMRE|autoMRE_IGN] [-j]
[-J MR|MR_DROP|MRE|STRICT|STRICT_DROP|T_<PERCENT>] [-k] [-K] [-L
MR|MRE|T_<PERCENT>] [-M] [-o outGroupName1[,outGroupName2[,...]]][-O]
[-p parsimonyRandomSeed] [-P proteinModel] [-q multipleModelFileName]
[-r binaryConstraintTree] [-R binaryModelParamFile] [-S
secondaryStructureFile] [-t userStartingTree] [-T numberOfThreads] [-u]
[-U] [-v] [-V] [-w outputDirectory] [-W slidingWindowSize] [-x
rapidBootstrapRandomNumberSeed] [-X] [-y] [-Y
quartetGroupingFileName|ancestralSequenceCandidatesFileName] [-z
multipleTreesFile] [-#|-N
numberOfRuns|autoFC|autoMR|autoMRE|autoMRE_IGN]
[--mesquite][--silent][--no-seq-check][--no-bfgs]
[--asc-corr=stamatakis|felsenstein|lewis]
[--flag-check][--auto-prot=ml|bic|aic|aicc]
[--epa-keep-placements=number][--epa-accumulated-threshold=threshold]
[--epa-prob-threshold=threshold] [--JC69][--K80][--HKY85]
[--bootstop-perms=number] [--quartets-without-replacement]
[---without-replacement] [--print-identical-sequences]

🎀 OPTIONS 🎀

  • -a — specify a column weight file name to assign individual weights to each column of the alignment. Those weights must be integers separated by any type and number of whitespaces whithin a separate file, see file "example_weights" for an example.
  • -A — specify one of the secondary structure substitution models implemented in RAxML. The same nomenclature as in the PHASE manual is used, available models: S6A, S6B, S6C, S6D, S6E, S7A, S7B, S7C, S7D, S7E, S7F, S16, S16A, S16B

DEFAULT: 16-state GTR model (S16)

  • -b — specify an integer number (random seed) and turn on bootstrapping

DEFAULT: OFF

  • -B — specify a floating point number between 0.0 and 1.0 that will be used as cutoff threshold for the MR-based bootstopping criteria. The recommended setting is 0.03.

DEFAULT: 0.03 (recommended empirically determined setting)

  • -c — specify number of distinct rate catgories for RAxML when model of rate heterogeneity is set to CAT Individual per-site rates are categorized into numberOfCategories rate categories to accelerate computations.

DEFAULT: 25

  • -C — enable verbose output for the "-L" and "-f i" options. This will produce more, as well as more verbose output files

DEFAULT: OFF

  • -d — start ML optimization from random starting tree

DEFAULT: OFF

  • -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 tree topology

DEFAULT: 0.1 for models not using proportion of invariant sites estimate

0.001 for models using proportion of invariant sites estimate

  • -E — specify an exclude file name, that contains a specification of alignment positions you wish to exclude. Format is similar to Nexus, the file shall contain entries like "100-200 300-400", to exclude a single column write, e.g., "100-100", if you use a mixed model, an appropriately adapted model file will be written.
  • -f — select algorithm:
    • -f a — rapid Bootstrap analysis and search for best-scoring ML tree in one program run
    • -f A — compute marginal ancestral states on a ROOTED reference tree provided with "-t"
    • -f b — draw bipartition information on a tree provided with "-t" based on multiple trees (e.g., from a bootstrap) in a file specified by "-z"
    • -f B — optimize br-len scaler and other model parameters (GTR, alpha, etc.) on a tree provided with "-t". The tree needs to contain branch lengths. The branch lengths will not be optimized, just scaled by a single common value.
    • -f c — check if the alignment can be properly read by RAxML
    • -f C — ancestral sequence test for Jiajie, users will also need to provide a list of taxon names via -Y separated by whitespaces
    • -f d — new rapid hill-climbing

DEFAULT: ON

  • -f D — rapid hill-climbing with RELL bootstraps
  • -f e — optimize model+branch lengths for given input tree under GAMMA/GAMMAI only
  • -f E — execute very fast experimental tree search, at present only for testing
  • -f F — execute fast experimental tree search, at present only for testing
  • -f g — compute per site log Likelihoods for one or more trees passed via -z and write them to a file that can be read by CONSEL. The model parameters will be estimated on the first tree only!
  • -f G — compute per site log Likelihoods for one or more trees passed via "-z" and write them to a file that can be read by CONSEL. The model parameters will be re-estimated for each tree
  • -f h — compute log likelihood test (SH-test) between best tree passed via "-t" and a bunch of other trees passed via "-z" The model parameters will be estimated on the first tree only!
  • -f H — compute log likelihood test (SH-test) between best tree passed via "-t" and a bunch of other trees passed via "-z" The model parameters will be re-estimated for each tree
  • -f i — calculate IC and TC scores (Salichos and Rokas 2013) on a tree provided with "-t" based on multiple trees (e.g., from a bootstrap) in a file specified by "-z"
  • -f I — a simple tree rooting algorithm for unrooted trees. It roots the tree by rooting it at the branch that best balances the subtree lengths (sum over branches in the subtrees) of the left and right subtree. A branch with an optimal balance does not always exist! You need to specify the tree you want to root via "-t".
  • -f j — generate a bunch of bootstrapped alignment files from an original alignment file. You need to specify a seed with "-b" and the number of replicates with "-#"
  • -f J — Compute SH-like support values on a given tree passed via "-t".
  • -f k — Fix long branch lengths in partitioned data sets with missing data using the branch length stealing algorithm. This option only works in conjunction with "-t", "-M", and "-q". It will print out a tree with shorter branch lengths, but having the same likelihood score.
  • -f m — compare bipartitions between two bunches of trees passed via "-t" and "-z" respectively. This will return the Pearson correlation between all bipartitions found in the two tree files. A file called RAxML_bipartitionFrequencies.outpuFileName will be printed that contains the pair-wise bipartition frequencies of the two sets
  • -f n — compute the log likelihood score of all trees contained in a tree file provided by "-z" under GAMMA or GAMMA+P-Invar. The model parameters will be estimated on the first tree only!
  • -f N — compute the log likelihood score of all trees contained in a tree file provided by "-z" under GAMMA or GAMMA+P-Invar. The model parameters will be re-estimated for each tree
  • -f o — old and slower rapid hill-climbing without heuristic cutoff
  • -f p — perform pure stepwise MP addition of new sequences to an incomplete starting tree and exit
  • -f P — perform a phylogenetic placement of sub trees specified in a file passed via "-z" into a given reference tree in which these subtrees are contained that is passed via "-t" using the evolutionary placement algorithm.
  • -f q — fast quartet calculator
  • -f r — compute pairwise Robinson-Foulds (RF) distances between all pairs of trees in a tree file passed via "-z" if the trees have node labales represented as integer support values the program will also compute two flavors of the weighted Robinson-Foulds (WRF) distance
  • -f R — compute all pairwise Robinson-Foulds (RF) distances between a large reference tree passed via "-t" and many smaller trees (that must have a subset of the taxa of the large tree) passed via "-z". This option is intended for checking the plausibility of very large phylogenies that can not be inspected visually any more.
  • -f s — split up a multi-gene partitioned alignment into the respective subalignments
  • -f S — compute site-specific placement bias using a leave one out test inspired by the evolutionary placement algorithm
  • -f t — do randomized tree searches on one fixed starting tree
  • -f T — do final thorough optimization of ML tree from rapid bootstrap search in stand-alone mode
  • -f u — execute morphological weight calibration using maximum likelihood, this will return a weight vector. you need to provide a morphological alignment and a reference tree via "-t"
  • -f v — classify a bunch of environmental sequences into a reference tree using thorough read insertions you will need to start RAxML with a non-comprehensive reference tree and an alignment containing all sequences (reference + query)
  • -f V — classify a bunch of environmental sequences into a reference tree using thorough read insertions you will need to start RAxML with a non-comprehensive reference tree and an alignment containing all sequences (reference + query)
WARNING: this is a test implementation for more efficient handling of multi-gene/whole-genome datasets!
  • -f w — compute ELW test on a bunch of trees passed via "-z". The model parameters will be estimated on the first tree only!
  • -f W — compute the ELW test on a bunch of trees you hand over via -z. The model parameters get re-estimated for each tree. Tedious for it, great for you 💅
  • -f x — compute pair-wise ML distances. The ML model parameters get estimated on either an MP starting tree or a user-defined tree passed via -t. Heads up bestie: this is only allowed for GAMMA-based models of rate heterogeneity.
  • -f y — classify a bunch of environmental sequences into a reference tree using parsimony. For this you'll need to start RAxML with a non-comprehensive reference tree and an alignment containing all the sequences (reference + query). The way that it needs both, fr fr.
DEFAULT for -f: new rapid hill climbing 🧗
  • -F — enable ML tree searches under the CAT model for very large trees without switching over to GAMMA at the end (saves memory, we love a low-memory queen). You can also use this option with the GAMMA models to skip that thorough optimization of the best-scoring ML tree at the end. No cap.

DEFAULT: OFF

  • -g — specify the file name of a multifurcating constraint tree. Plot twist: this tree does not need to be comprehensive, i.e. it must not contain all taxa. She's flexible like that ✨
  • -G — enable the ML-based evolutionary placement algorithm heuristics by specifying a threshold value — that's the fraction of insertion branches to be evaluated using slow insertions under ML. 🦋
  • -h — display this help message. Your SOS button, bestie 💖
  • -H — disable pattern compression.

DEFAULT: ON

  • -i — initial rearrangement setting for the subsequent application of the topological changes phase. Shake things up a little 🎀
  • -I — a posteriori bootstopping analysis. Use one of these:
    • -I autoFC — for the frequency-based criterion
    • -I autoMR — for the majority-rule consensus tree criterion
    • -I autoMRE — for the extended majority-rule consensus tree criterion
    • -I autoMRE_IGN — for metrics similar to MRE, but it includes bipartitions under the threshold whether they're compatible or not. This emulates MRE but is faster to compute. ⚡

You also need to pass a tree file containing several bootstrap replicates via -z. Don't forget it, she will not work without it 😭

  • -j — specifies that intermediate tree files shall be written to file during the standard ML and BS tree searches. Little progress pics along the way 📸

DEFAULT: OFF

  • -J — compute a majority rule consensus tree with -J MR, or an extended majority rule consensus tree with -J MRE, or a strict consensus tree with -J STRICT. For a custom consensus threshold that's >= 50%, specify T_<NUM>, where 100 >= NUM >= 50. And the options -J STRICT_DROP and -J MR_DROP will execute an algorithm that identifies dropsets containing rogue taxa, as proposed by Pattengale et al. in the paper "Uncovering hidden phylogenetic consensus". You'll also need to provide a tree file containing several UNROOTED trees via -z. Rooted? We don't know her 💀
  • -k — specifies that bootstrapped trees should be printed with branch lengths. Plot twist: the bootstraps run a bit longer, because the model parameters get optimized at the end of each run under GAMMA or GAMMA+P-Invar respectively. Worth it though 💜

DEFAULT: OFF

  • -K — specify one of the multi-state substitution models (max 32 states) implemented in RAxML. The available models are: ORDERED, MK, GTR.

DEFAULT: GTR model

  • -L — compute consensus trees labelled by IC supports and the overall TC value, as proposed in Salichos and Rokas 2013. Compute a majority rule consensus tree with -L MR, or an extended majority rule consensus tree with -L MRE. For a custom consensus threshold that's >= 50%, specify -L T_<NUM>, where 100 >= NUM >= 50. And you will of course also need to provide a tree file containing several UNROOTED trees via -z! 👑
  • -m — model of Binary (Morphological), Nucleotide, Multi-State, or Amino Acid Substitution. This is the main character of the whole option list, so buckle up 🌈

BINARY

  • -m BINCAT[X] — optimization of site-specific evolutionary rates which are categorized into numberOfCategories distinct rate categories, for greater computational efficiency. The final tree might be evaluated automatically under BINGAMMA, depending on the tree search option. With the optional X appendix you can specify a ML estimate of base frequencies.
  • -m BINCATI[X] — optimization of site-specific evolutionary rates which are categorized into numberOfCategories distinct rate categories, for greater computational efficiency. The final tree might be evaluated automatically under BINGAMMAI, depending on the tree search option. With the optional X appendix you can specify a ML estimate of base frequencies.
  • -m ASC_BINCAT[X] — optimization of site-specific evolutionary rates which are categorized into numberOfCategories distinct rate categories, for greater computational efficiency. The final tree might be evaluated automatically under BINGAMMA, depending on the tree search option. With the optional X appendix you can specify a ML estimate of base frequencies. The ASC prefix will correct the likelihood for ascertainment bias. ✨
  • -m BINGAMMA[X] — GAMMA model of rate heterogeneity (the alpha parameter will be estimated). With the optional X appendix you can specify a ML estimate of base frequencies.
  • -m ASC_BINGAMMA[X] — GAMMA model of rate heterogeneity (the alpha parameter will be estimated). The ASC prefix will correct the likelihood for ascertainment bias. With the optional X appendix you can specify a ML estimate of base frequencies.
  • -m BINGAMMAI[X] — same as BINGAMMA, but with an estimate of the proportion of invariable sites. With the optional X appendix you can specify a ML estimate of base frequencies.

NUCLEOTIDES

  • -m GTRCAT[X] — GTR + optimization of substitution rates + optimization of site-specific evolutionary rates which are categorized into numberOfCategories distinct rate categories, for greater computational efficiency. The final tree might be evaluated under GTRGAMMA, depending on the tree search option. With the optional X appendix you can specify a ML estimate of base frequencies.
  • -m GTRCATI[X] — GTR + optimization of substitution rates + optimization of site-specific evolutionary rates which are categorized into numberOfCategories distinct rate categories, for greater computational efficiency. The final tree might be evaluated under GTRGAMMAI, depending on the tree search option. With the optional X appendix you can specify a ML estimate of base frequencies.
  • -m ASC_GTRCAT[X] — GTR + optimization of substitution rates + optimization of site-specific evolutionary rates which are categorized into numberOfCategories distinct rate categories, for greater computational efficiency. The final tree might be evaluated under GTRGAMMA, depending on the tree search option. With the optional X appendix you can specify a ML estimate of base frequencies. The ASC prefix will correct the likelihood for ascertainment bias. 🦋
  • -m GTRGAMMA[X] — GTR + optimization of substitution rates + GAMMA model of rate heterogeneity (the alpha parameter will be estimated). With the optional X appendix you can specify a ML estimate of base frequencies.
  • -m ASC_GTRGAMMA[X] — GTR + optimization of substitution rates + GAMMA model of rate heterogeneity (the alpha parameter will be estimated). The ASC prefix will correct the likelihood for ascertainment bias. With the optional X appendix you can specify a ML estimate of base frequencies.
  • -m GTRGAMMAI[X] — same as GTRGAMMA, but with an estimate of the proportion of invariable sites. With the optional X appendix you can specify a ML estimate of base frequencies.

MULTI-STATE

  • -m MULTICAT[X] — optimization of site-specific evolutionary rates which are categorized into numberOfCategories distinct rate categories, for greater computational efficiency. The final tree might be evaluated automatically under MULTIGAMMA, depending on the tree search option. With the optional X appendix you can specify a ML estimate of base frequencies.
  • -m MULTICATI[X] — optimization of site-specific evolutionary rates which are categorized into numberOfCategories distinct rate categories, for greater computational efficiency. The final tree might be evaluated automatically under MULTIGAMMAI, depending on the tree search option. With the optional X appendix you can specify a ML estimate of base frequencies.
  • -m ASC_MULTICAT[X] — optimization of site-specific evolutionary rates which are categorized into numberOfCategories distinct rate categories, for greater computational efficiency. The final tree might be evaluated automatically under MULTIGAMMA, depending on the tree search option. With the optional X appendix you can specify a ML estimate of base frequencies. The ASC prefix will correct the likelihood for ascertainment bias. 💅
  • -m MULTIGAMMA[X] — GAMMA model of rate heterogeneity (the alpha parameter will be estimated). With the optional X appendix you can specify a ML estimate of base frequencies.
  • -m ASC_MULTIGAMMA[X] — GAMMA model of rate heterogeneity (the alpha parameter will be estimated). The ASC prefix will correct the likelihood for ascertainment bias. With the optional X appendix you can specify a ML estimate of base frequencies.
  • -m MULTIGAMMAI[X] — same as MULTIGAMMA, but with an estimate of the proportion of invariable sites. With the optional X appendix you can specify a ML estimate of base frequencies.

Okay so, multi-state tea 🫖 — you can use up to 32 distinct character states to encode multi-state regions, and they must be used in this exact order:

0, 1, 2, 3, 4, 5, 6, 7, 8, 9, A, B, C, D, E, F, G, H, I, J, K, L, M,
N, O, P, Q, R, S, T, U, V

i.e., if you have 6 distinct character states you would use 0, 1, 2, 3, 4, 5 to encode them. The substitution model for the multi-state regions can be selected via the -K option. Slay ✨

AMINO ACIDS

  • -m PROTCATmatrixName[F|X] — specified AA matrix + optimization of substitution rates + optimization of site-specific
  • -m PROTCATImatrixName[F|X] — same energy as PROTCATmatrixName[F|X] but with a capital I doing the most: you pick that AA matrix, you optimize the substitution rates, and you optimize the site-specific evolutionary rates too, which get sorted into numberOfCategories distinct rate categories so the whole thing doesn't take a million years. The final tree might get evaluated automatically under PROTGAMMAImatrixName[F|X], depends on your tree search option. Optional "X" on the end and you get a ML estimate of base frequencies. 💅
  • -m ASC_PROTCATmatrixName[F|X] — specified AA matrix + optimization of substitution rates + optimization of site-specific evolutionary rates, categorized into numberOfCategories distinct rate categories for that sweet computational efficiency. Final tree might be auto-evaluated under PROTGAMMAmatrixName[F|X], depending on the tree search option. Optional "X" appendix → ML estimate of base frequencies. And the ASC prefix? It corrects the likelihood for ascertainment bias. no cap, it's just doing its job.
  • -m PROTGAMMAmatrixName[F|X] — specified AA matrix + optimization of substitution rates + the GAMMA model of rate heterogeneity (the alpha parameter will be estimated, she's doing math). Optional "X" appendix → ML estimate of base frequencies. ✨
  • -m ASC_PROTGAMMAmatrixName[F|X] — specified AA matrix + optimization of substitution rates + GAMMA model of rate heterogeneity (yep, alpha parameter gets estimated). The ASC prefix will correct the likelihood for ascertainment bias. Optional "X" appendix → ML estimate of base frequencies.
  • -m PROTGAMMAImatrixName[F|X] — literally the same as PROTGAMMAmatrixName[F|X], except it also estimates the proportion of invariable sites. Optional "X" appendix → ML estimate of base frequencies, as always.
Bestie, here's the full menu of AA substitution models you can drop in for matrixName: DAYHOFF, DCMUT, JTT, MTREV, WAG, RTREV, CPREV, VT, BLOSUM62, MTMAM, LG, MTART, MTZOA, PMB, HIVB, HIVW, JTTDCMUT, FLU, STMTREV, DUMMY, DUMMY2, AUTO, LG4M, LG4X, PROT_FILE, GTR_UNLINKED, GTR.

Throw on the optional "F" appendix and you're saying you want empirical base frequencies. Heads up though: AUTOF and AUTOX are not supported any more — so if you specify AUTO, it will now test prot subst. models with and without empirical base frequencies. Also, for partitioned models you can additionally specify the per-gene AA model in the partition file (peep the manual for the details). And one more thing: if you estimate AA GTR parameters on a partitioned dataset, they get linked (estimated jointly) across all partitions, to avoid over-parametrization. team work makes the dream work 🫶

  • -M — turns on estimation of individual per-partition branch lengths. Only vibes when used with -q. Branch lengths for individual partitions get printed to separate files, and a weighted average of the branch lengths is computed using the respective partition lengths.

DEFAULT: OFF

  • -n — the name of the output file. that's it, that's the job.
  • -o — name a single outgroup, or a comma-separated list of outgroups, e.g. -o Rat or -o Rat,Mouse. If the multiple outgroups turn out not to be monophyletic, the first name in the list gets picked as the outgroup. And do not leave spaces between taxon names!
  • -O — disables the check for a completely undetermined sequence in the alignment. With -O specified, the program will not exit with an error message when it finds one.

DEFAULT: check enabled

  • -p — a random number seed for the parsimony inferences. This lets you reproduce your results, and it helps the author debug the program. 🎀
  • -P — file name of a user-defined AA (Protein) substitution model. This file must contain 420 entries: the first 400 are the AA substitution rates (and this must be a symmetric matrix), and the last 20 are the empirical base frequencies.
  • -q — file name of the file holding the assignment of models to alignment partitions, for multiple models of substitution. For the syntax of that file, go consult the manual. 📖
  • -r — file name of a binary constraint tree. This tree does not need to be comprehensive, i.e. it must not contain all taxa.
  • -R — file name of a binary model parameter file that was previously generated with RAxML using the -f e tree evaluation option. The file name should look like: RAxML_binaryModelParameters.runID
  • -s — the name of the alignment data file, in PHYLIP format.
  • -S — the name of a secondary structure file. The file can contain "." for alignment columns that do not form part of a stem, plus the characters "()<>[]{}" to define stem regions and pseudoknots. 🦋
  • -t — a user starting tree file name, in Newick format.
  • -T — PTHREADS VERSION ONLY! The number of threads you want to run. Make sure you set -T to at most the number of CPUs you have on your machine — otherwise there will be a huge performance decrease! 😭 like bestie I am begging you, count your cores first.
  • -u — use the median for the discrete approximation of the GAMMA model of rate heterogeneity.

DEFAULT: OFF

  • -U — tries to save memory by using a SEV-based implementation for gap columns on large gappy alignments. The technique is described here: http://www.biomedcentral.com/1471-2105/12/470. This will only work for DNA and/or PROTEIN data, and only with the SSE3 or AVX-vextorized version of the code. 💀
  • -v — display version information.
  • -V — disables the rate heterogeneity among sites model and uses one without rate heterogeneity instead. Only works if you specified the CAT model of rate heterogeneity.

DEFAULT: use rate heterogeneity

  • -w — the FULL (!) path to the directory into which RAxML shall write its output files. 👑

DEFAULT: current directory

  • -W — sliding window size for the leave-one-out site-specific placement bias algorithm. Only effective when used together with -f S.

DEFAULT: 100 sites

  • -x — an integer number (random seed) that turns on rapid bootstrapping. CAUTION: unlike in version 7.0.4, RAxML will conduct rapid BS replicates under the model of rate heterogeneity you specified via -m, and not by default under CAT.
  • -X — same as the -y option below, however the parsimony search is more superficial. RAxML will only do a randomized stepwise addition order parsimony tree reconstruction, without performing any additional SPRs. This may be helpful for very broad whole-genome datasets, since it can generate topologically more different starting trees.

DEFAULT: OFF

  • -y — if you want to only compute a parsimony starting tree with RAxML, specify -y and the program will exit after computing the starting tree.

DEFAULT: OFF

  • -Y — pass a quartet grouping file name that defines four groups from which to draw quartets. The file input format must contain 4 groups, in this form: (Chicken, Human, Loach), (Cow, Carp), (Mouse, Rat, Seal), (Whale, Frog); Only works in combination with -f q!
  • -z — the file name of a file containing multiple trees, e.g. from a bootstrap, that shall be used to draw bipartition values onto a tree provided with -t. It can also be used to compute per-site log likelihoods in combination with -f g, and to read a bunch of trees for a couple of other options (-f h, -f m, -f n). versatile queen 🌈
  • -#|-N — the number of alternative runs on distinct starting trees. Combined with the -b option, this will invoke a multiple bootstrap analysis. Note that -N was added as an alternative because -# sometimes caused problems with certain MPI job submission systems, since -# is often used to start comments. 🫠 If you want to use the bootstopping criteria, specify -# autoMR or -# autoMRE or -# autoMRE_IGN for the majority-rule tree based criteria (see -I option), or -# autoFC for the frequency-based criterion. Bootstopping will only work in combination with -x or -b.

DEFAULT: 1 single analysis

  • --mesquite — print output files that can be parsed by Mesquite.

DEFAULT: Off

  • --silent — disables printout of warnings related to identical sequences and entirely undetermined sites in the alignment.

DEFAULT: Off

  • --no-seq-check — disables checking the input MSA for identical sequences and entirely undetermined sites. Enabling this option may save time, in particular for large phylogenomic alignments. But before you use it, make sure to check the alignment using the -f c option! 🚨

DEFAULT: Off

  • --no-bfgs — disables the automatic usage of the BFGS method to optimize GTR rates on unpartitioned DNA datasets.

DEFAULT: BFGS on

  • --asc-corr — specify the type of ascertainment bias correction you want to use. There are 3 types available:
    • --asc-corr=lewis: the standard correction by Paul Lewis.
    • --asc-corr=felsenstein: a correction introduced by Joe Felsenstein that allows you to explicitly specify the number of invariable sites (if known) you want to correct for.
    • --asc-corr=stamatakis: a correction introduced by the author himself that allows you to explicitly specify the number of invariable sites for each character (if known) you want to correct for. 💅
  • --flag-check — when you use this option, RAxML will only check whether all the command line flags specified are available, then exit with a message listing all invalid command line flags, or with a message stating that all flags are valid. vibe check for your flags ✅
  • --auto-prot=ml|bic|aic|aicc — when using automatic protein model selection, you can choose 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

  • --epa-keep-placements=number — tell it how many potential placements you wanna keep for each read in the EPA algorithm, bestie.
    • DEFAULT: 7
heads up though: the actual values printed also depend on whatever you set for --epa-prob-threshold=threshold ! so don't come for me if the output isn't the number you typed 💀
  • --epa-prob-threshold=threshold — a percent threshold for including potential placements of a read, based on the maximum placement weight for that read. So like, if you set this to 0.01, placements whose placement weight is 1 per cent of the maximum placement will still be printed to file — but only if the setting of --epa-keep-placements allows for it.
    • DEFAULT: 0.01
  • --epa-accumulated-threshold=threshold — an accumulated likelihood weight threshold for which different placements of a read get printed to file. Placements for a read keep getting printed until the sum of their placement weights has reached the threshold value.
lowkey important: this option can neither be used in combination with --epa-prob-threshold nor with --epa-keep-placements! pick your fighter, not both 🙈
  • --JC69 — say that all DNA partitions will evolve under the Jukes-Cantor model. This overrides all other model specifications for DNA partitions.
    • DEFAULT: Off
  • --K80 — say that all DNA partitions will evolve under the K80 model. This overrides all other model specifications for DNA partitions.
    • DEFAULT: Off
  • --HKY85 — say that all DNA partitions will evolve under the HKY85 model. This overrides all other model specifications for DNA partitions.
    • DEFAULT: Off
  • --bootstop-perms=number — how many permutations you want conducted for the bootstopping / bootstrap convergence test.
the allowed minimum number is 100! fr fr, don't try to be cute with a smaller number.
  • DEFAULT: 100
  • --quartets-without-replacement — quartets get randomly subsampled, but without replacement.
    • DEFAULT: random sampling with replacements
  • --print-identical-sequences — RAxML shall automatically generate a .reduced alignment with all undetermined columns removed, but without removing exactly identical sequences.
    • DEFAULT: identical sequences will also be removed in the .reduced file

See also 💬

Please also consult the RAxML-manual. It's giving required reading, ngl.

Please report bugs via the RAxML google group! 🐛 And when you do, send us all of it: all input files, the exact invocation, details of the HW and operating system, as well as all error messages printed to screen. The more receipts the better, iykyk.

👑 AUTHOR 👑

This manpage was written by Andreas Tille for the Debian distribution and can be used for any other usage of the program. 💖

The code itself was written by Alexandros Stamatakis.

With greatly appreciated code contributions by:

  • Andre Aberer (HITS)
  • Simon Berger (HITS)
  • Alexey Kozlov (HITS)
  • Kassian Kobert (HITS)
  • David Dao (KIT and HITS)
  • Sarah Lutteropp (KIT and HITS)
  • Nick Pattengale (Sandia)
  • Wayne Pfeiffer (SDSC)
  • Akifumi S. Tanabe (NRIFS)
  • Charlie Taylor (UF)

and that's the whole page bestie 💅 this one's from raxmlHPC 8.2.12, dated December 2019 — go build some trees, you absolute icon 🦋🌸

📦 where this page comes from