From de15ccdcb7be683af01c42544c78a8bdad798df6 Mon Sep 17 00:00:00 2001 From: Siavash Mirarab Date: Fri, 17 Jun 2016 09:42:11 -0500 Subject: [PATCH 1/5] Update astral-tutorial.md --- astral-tutorial.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/astral-tutorial.md b/astral-tutorial.md index 0982e3727..04cf90ad9 100644 --- a/astral-tutorial.md +++ b/astral-tutorial.md @@ -122,7 +122,7 @@ A larger real dataset from the [1kp](http://www.pnas.org/content/early/2014/10/2 424 genes from 103 species. Run: ``` -java -jar astral.4.10.6.jar -i test_data/1KP-genetrees.tre -o test_data/1kp-tre +java -jar astral.4.10.6.jar -i test_data/1KP-genetrees.tre -o test_data/1kp.tre ``` This takes about a minute to run on a powerful laptop. On this dataset, notice in the ASTRAL log information that it originally starts with 11043 clusters in its search space, and using heuristics implemented in ASTRAL-II, it increases the search space slightly to 11085 clusters. For more challenging datasets (i.e., more discordance or fewer genes) this number might increase a lot. @@ -181,7 +181,7 @@ Here is a description of various information that can be turned on by using `-t` Run: ``` -java -jar astral.4.10.6.jar -q test_data/simulated_14taxon.default.tre -i test_data/simulated_14taxon.gene.tre -t 2 +java -jar astral.4.10.6.jar -q test_data/1kp.tre -i test_data/1KP-genetrees.tre -t 2 -o test_data/1kp-scored.tre ``` read all the values given for a couple of branches and try to make sense of them. @@ -357,4 +357,4 @@ ASTRAL code uses bytecode and some reverse engineered code from PhyloNet package ### Bug Reports: -contact: ``astral-users@googlegroups.com`` \ No newline at end of file +contact: ``astral-users@googlegroups.com`` From 5a93c4c2a049330e5e82c729386763ae3c20ce97 Mon Sep 17 00:00:00 2001 From: Siavash Mirarab Date: Fri, 17 Jun 2016 09:43:06 -0500 Subject: [PATCH 2/5] Update astral-tutorial.md --- astral-tutorial.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/astral-tutorial.md b/astral-tutorial.md index 04cf90ad9..c6df1a439 100644 --- a/astral-tutorial.md +++ b/astral-tutorial.md @@ -186,7 +186,7 @@ java -jar astral.4.10.6.jar -q test_data/1kp.tre -i test_data/1KP-genetrees.tre read all the values given for a couple of branches and try to make sense of them. -#### Priori hyper-parameter +#### Prior hyper-parameter Our calculations of the local posterior probabilities and branch lengths use a Yule prior model for the branch lengths of the species tree. The speciation rate (in coalescent units) of the Yule process (lambda) is by default set to 0.5, which results in a flat prior for the quartet frequencies in the `[1/3,1]` range. Using `-c` option one can adjust the hyper-parameter for the prior. For example, you might want to estimate lambda from the data after one run and plug the estimate prior in a subsequent run. We have not yet fully explored the impact of lambda on the posterior. For branch lengths, lambda acts as a pseudocount and can have a substantial impact on the estimated branch length for very long branches. More specifically, if there is no, or very little discordance around a branch, the MAP lengths of the branch (which is what we report) is almost fully determined by the prior. From 561b886cc1d90d3fe3cc95d02d6b0d5fc229eafb Mon Sep 17 00:00:00 2001 From: Siavash Mirarab Date: Fri, 17 Jun 2016 10:59:12 -0500 Subject: [PATCH 3/5] Update astral-tutorial.md --- astral-tutorial.md | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/astral-tutorial.md b/astral-tutorial.md index c6df1a439..2972c5c73 100644 --- a/astral-tutorial.md +++ b/astral-tutorial.md @@ -176,12 +176,18 @@ Here is a description of various information that can be turned on by using `-t` * `pp1`, `pp2`, `pp3`: these three show the local posterior probabilities (as defined in the description of `-t 4`) for the main topology, the first alternative, and the second alternative, respectively. * `QC`: this shows the total number of quartets defined around each branch (this is what our paper calls `m`). * `EN`: this is the effective number of genes for the branch. If you don't have any missing data, this would be the number of branches in your tree. When there is missing data, some gene trees might have nothing to say about a branch. Thus, the effective number of genes might be smaller than the total number of genes. - +* *Alternative quartet topologies* (`-t 8`): Outputs `q1`,`q2`,`q3`; these three values show quartet support (as defined in the description of `-t 1`) for the main topology, the first alternative, and the second alternative, respectively. Run: ``` -java -jar astral.4.10.6.jar -q test_data/1kp.tre -i test_data/1KP-genetrees.tre -t 2 -o test_data/1kp-scored.tre +java -jar astral.4.10.6.jar -q test_data/1kp.tre -i test_data/1KP-genetrees.tre -t 2 -o test_data/1kp-scored-t2.tre +``` +``` +java -jar astral.4.10.6.jar -q test_data/1kp.tre -i test_data/1KP-genetrees.tre -t 4 -o test_data/1kp-scored-t4.tre +``` +``` +java -jar astral.4.10.6.jar -q test_data/1kp.tre -i test_data/1KP-genetrees.tre -t 8 -o test_data/1kp-scored-t8.tre ``` read all the values given for a couple of branches and try to make sense of them. From 4e2533e29370f6947b7f08650c221e84370150b4 Mon Sep 17 00:00:00 2001 From: Siavash Mirarab Date: Fri, 17 Jun 2016 11:04:58 -0500 Subject: [PATCH 4/5] Update astral-tutorial.md --- astral-tutorial.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/astral-tutorial.md b/astral-tutorial.md index 2972c5c73..38665adb9 100644 --- a/astral-tutorial.md +++ b/astral-tutorial.md @@ -210,7 +210,7 @@ Note that setting lambda to 0 results in reporting ML estimates of the branch le ### Multi-locus Bootstrapping: -Recent versions of ASTRAL output a branch support value even without bootstrapping. Our [analuses]](http://mbe.oxfordjournals.org/content/early/2016/05/12/molbev.msw079.short?rss=1) have revealed that this form of support is more reliable than bootstrapping under conditions we have compared the two methods. Nevertheless, you may want to run bootstrapping as well. +Recent versions of ASTRAL output a branch support value even without bootstrapping. Our [analyses](http://mbe.oxfordjournals.org/content/early/2016/05/12/molbev.msw079.short?rss=1) have revealed that this form of support is more reliable than bootstrapping under conditions we have compared the two methods. Nevertheless, you may want to run bootstrapping as well. Astral can perform multi-locus bootstrapping ([Seo, 2008](http://www.ncbi.nlm.nih.gov/pubmed/18281270)). To be able to perform multi-locus bootstrapping, ASTRAL needs to have access to bootstrap replicates for each gene. To start multi-locus bootstrapping using ASTRAL, you need to provide the location of all gene tree bootstrap replicates. To run bootstrapping on our test input files, From a985f96f75d680893868c7ebe30628030d8acd07 Mon Sep 17 00:00:00 2001 From: Siavash Mirarab Date: Fri, 17 Jun 2016 11:31:41 -0500 Subject: [PATCH 5/5] Update astral-tutorial.md --- astral-tutorial.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/astral-tutorial.md b/astral-tutorial.md index 38665adb9..b4cb4c071 100644 --- a/astral-tutorial.md +++ b/astral-tutorial.md @@ -199,11 +199,11 @@ Our calculations of the local posterior probabilities and branch lengths use a Y Run the following two commands and compare the lengths of the longest branches: ``` -java -jar astral.4.10.6.jar -q test_data/simulated_14taxon.default.tre -i test_data/simulated_14taxon.gene.tre -c 0.001 +java -jar astral.4.10.6.jar -q test_data/1kp.tre -i test_data/1KP-genetrees.tre -c 2 -o test_data/1kp-scored-c2.tre ``` ``` -java -jar astral.4.10.6.jar -q test_data/simulated_14taxon.default.tre -i test_data/simulated_14taxon.gene.tre -c 2 +java -jar astral.4.10.6.jar -q test_data/1kp.tre -i test_data/1KP-genetrees.tre -c 0.001 -o test_data/1kp-scored-cs.tre ``` Note that setting lambda to 0 results in reporting ML estimates of the branch lengths instead of MAP. However, for branches with no discordance, we cannot compute a branch lengths. For these, we currently arbitrarily set ML to 10 coalescent units (we might change this in future versions).