Segmentation fault: 11 for -f k

[Ana Jesovnik]

I am getting the "Segmentation fault: 11", even though i should have enough memory (according to RaxML calculator) to run the following command:

raxmlHPC -f k -m GTRCAT -s mafft-gb-90p-miss-taxa.phylip -q serico_partitions90.txt -M -t RAxML_bipartitions.serico_90p_raxml_best_tre_rename -n fk_serico_90p

My matrix is 93 taxa x 702574 bp

I have the latest version of RAxMl installed, RAxML version 8.2.8

Thank you for your time!

[Alexandros Stamatakis]

how much memory do you have and how much did you calculate?

the calculator gives you a lower bound, realistically raxml will need
about twice the memory that is being calculated by the on-line tool ...

if you have a lot of missing data you can use the memory saving option:

-U this will save memory proportional to the amount of missing data in
your MSA,

alexis

> --
> You received this message because you are subscribed to the Google
> Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to raxml+un...@googlegroups.com
> <mailto:raxml+un...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.

--
Alexandros (Alexis) Stamatakis

Research Group Leader, Heidelberg Institute for Theoretical Studies
Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology
Adjunct Professor, Dept. of Ecology and Evolutionary Biology, University
of Arizona at Tucson

www.exelixis-lab.org

[Ted Schultz]

I just tried -f k using the -U command and got segmentation fault 11.  Has anyone successfully run -f k?  I would really like to try it.

[Alexandros Stamatakis]

Dear Ted,

This option won't work on conjunction with -U.

Also note that there is a stand-alone implementation for this procedure
described in this paper here:

http://bioinformatics.oxfordjournals.org/content/32/9/1331

the code is available at:

https://github.com/ddarriba/ForeSeqs

Alexis

> > an email to raxml+un...@googlegroups.com <javascript:>
> > <mailto:raxml+un...@googlegroups.com <javascript:>>.

> > For more options, visit https://groups.google.com/d/optout

> <https://groups.google.com/d/optout>.

>
> --
> Alexandros (Alexis) Stamatakis
>
> Research Group Leader, Heidelberg Institute for Theoretical Studies
> Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology
> Adjunct Professor, Dept. of Ecology and Evolutionary Biology,
> University
> of Arizona at Tucson
>

> www.exelixis-lab.org <http://www.exelixis-lab.org>

[Ted Schultz]

Alexis:

Thanks for your reply!  Perhaps my best bet is to try the stand-alone.  However, due to my partitions, I may encounter the same problem described below.

In trying to work with the -f k command, I am running into a problem with partitioning.  I am using the partition definitions created by PartitionFinder.  They work perfectly well with the -q command in a RAxML tree search.  But when I use them with the -f k command, it is clear that the partitions are not properly recognized.  It is probably important to note that each of the troublesome partitions (p1, p2, p3) includes non-contiguous sites (e.g., codon positions 1 and 2 but not 3).  I am on a Mac (3.5 GHZ 6-CORE INTEL XEON E5, OSX V.10.9.5, 64GB RAM).  I am using SSE3 RAxML 8.2.3.  Here is the relevant output:

Alignment has 1385 columns

Found 0 invariant alignment patterns that correspond to 0 columns 

Proportion of gaps and completely undetermined characters in this alignment: 40.57%

RAxML branch length stealing

Using 4 distinct models/data partitions with individual per partition branch length optimization

All free model parameters will be estimated by RAxML

GAMMA+P-Invar model of rate heteorgeneity, ML estimate of alpha-parameter

GAMMA+P-Invar Model parameters will be estimated up to an accuracy of 0.0010000000 Log Likelihood units

Partition: 0

Alignment Patterns: 2

Name: p1

DataType: DNA

Substitution Matrix: GTR

Partition: 1

Alignment Patterns: 1

Name: p2

DataType: DNA

Substitution Matrix: GTR

Partition: 2

Alignment Patterns: 2

Name: p3

DataType: DNA

Substitution Matrix: GTR

Partition: 3

Alignment Patterns: 1380

Name: p4

DataType: DNA

Substitution Matrix: GTR

RAxML was called as follows:

raxmlHPC-SSE3 -f k -m GTRGAMMAI -s FUN153x1385.phy -q FUN153_parts.txt -M -t RAxML_bestTree.FUN153x1385_RAxML.out.tre -n FUN153x1385_RAxML_BLrecalc.out 

Partition p1 number 0 has a problem, the number of expected states is 4 the number of states that are present is 2.

Please go and fix your data!

Partition p2 number 1 has a problem, the number of expected states is 4 the number of states that are present is 1.

Please go and fix your data!

Partition p3 number 2 has a problem, the number of expected states is 4 the number of states that are present is 2.

Please go and fix your data!

raxmlHPC-SSE3(327,0x7fff7d5f5310) malloc: *** error for object 0x7f9458702088: incorrect checksum for freed object - object was probably modified after being freed.

*** set a breakpoint in malloc_error_break to debug

Abort trap: 6

[Alexandros Stamatakis]

Dear Ted,

> Thanks for your reply! Perhaps my best bet is to try the stand-alone.

Yes please, it's more comprehensive than what we have in RAxML, the
RAxMl option for this should actually be deprecated.

> However, due to my partitions, I may encounter the same problem
> described below.
>
> In trying to work with the -f k command, I am running into a problem
> with partitioning. I am using the partition definitions created by
> PartitionFinder. They work perfectly well with the -q command in a
> RAxML tree search. But when I use them with the -f k command, it is
> clear that the partitions are not properly recognized. It is probably
> important to note that each of the troublesome partitions (p1, p2, p3)
> includes non-contiguous sites (e.g., codon positions 1 and 2 but not 3).
> I am on a Mac (3.5 GHZ 6-CORE INTEL XEON E5, OSX V.10.9.5, 64GB RAM).
> I am using SSE3 RAxML 8.2.3. Here is the relevant output:

Okay, note that the latest version is 8.2.8 also note those very
important warnings raxml is giving you for your partitions:

"Partition p1 number 0 has a problem, the number of expected states is 4
the number of states that are present is 2. Please go and fix your data!"

Please see previous discussions on this topic on this google group.

Alexis

[Ted Schultz]

Alexis:

I tried an experiment.  I moved sites around so that all the partitions were contiguous.  I didn't change anything else except to update the partition file to reflect the new, contiguous partitions.  With those changes, the branch-length-stealing algorithm worked.  So I'm pretty sure the RAxML -f k command doesn't like non-contiguous partitions, at least in v.8.2.3.

Ted

[Alexandros Stamatakis]

hi ted,

yes that might well be, but as I said this option needs to be remnoved
from raxml anyway, because the implementation is incomplete compared to
what we describe in the paper ...

thanks for checking ...

alexis