[wgs-assembler-users] fragmented assemblies with CA 8.3rc2?

[A. B. C. <ber...@gm...>]

Dear all,
I installed the CA 8.3rc2 in my server (Dell PE2900 running CentOS6.6,
with 8 cores and 64 Gb RAM) . As I have not used CA before, I run two
sets of test data from the PBcR web pages (or from the MHAP paper).
In both cases my assembly was quite more fragmented then the reported
ones. The datasets are:


1)  30x coverage E. coli (
http://wgs-assembler.sourceforge.net/wiki/index.php/PBcR#Assembling_an_E._coli
). It should have assembled as a single contig, and I always got 11
contigs. My command line and spec file follows:

PBcR -pbCNS -length 500  -l K12_attempt11 -s
/home/tools/CA_tests/yeast/yeast2.spec  -fastq
/home/tools/CA_tests/ecoli_83/selfSampleData/pacbio_filtered.fastq
genomeSize=4650000 > run11.out 2>&1 &

spec file:
useGrid = 0
scriptOnGrid = 0
assemble = 1
javaPath=/usr/bin/
ovlThreads = 12
threads = 12
ovlConcurrency = 1
cnsConcurrency = 12
merylThreads = 12

I also removed the last 5 lines of the spec file (allowing PBcR to
choose nearly everything), but againg got 11 contigs. However, if I
run the program with the full Ecoli data set (downloaded from the AWS
snapshot), I got a single contig.



2) The yeast data set reported in the MHAP paper (Berlin et al 2015) ,
downloaded from
http://gembox.cbcb.umd.edu/mhap/raw/yeast_filtered.fastq.gz
The MHAP paper describes that the assembly resulted in 21 contigs,
whereas I am always getting around 30. The command line follows (the
spec file is the same used for Ecoli ):

PBcR   -length 500  -l yeast2 -s yeast2.spec  -fastq
/home/tools/CA_tests/yeast/yeast_filtered.fastq genomeSize=12100000 >
yeast2.out 2>&1 &

I also tried to force the use of PBDAGCON, instead of falcon_sense, by
adding the line " pbcns=1" to the spec file, , and removing the
-pbCNS from the command line. The assembly was much slower, and, to my
surprise, more fragmented: 39 contigs.

Are these  results normal, or do they indicate some problem in my
installation of the Celera Assembler?

Yours,
Bernardo


A. Bernardo Carvalho

Departamento de Genética
Universidade Federal do Rio de Janeiro