BayesCall -- has anyone tried it?

[jnfass]

Has anyone had success trying out BayesCall? I'm wondering if it's worth benchmarking ...

Genome Res. 2009 Oct;19(10):1884-95. Epub 2009 Aug 6.
BayesCall: A model-based base-calling algorithm for high-throughput short-read sequencing.
Kao WC, Stevens K, Song YS.

(available here):

http://bayescall.sourceforge.net/

[mehdik]

I installed BayesCall, I am able to install all the prerequisites as well as BayesCall. The first part EM.py starts but I am stuck at multiple runtime erros in the python scripts. Anyone succeeded in properly running the program?

[darked89]

BayesCall -- has anyone tried it?

I installed BayesCall, I am able to install all the prerequisites as well as BayesCall. The first part EM.py starts but I am stuck at multiple runtime errors in the python scripts. Anyone succeeded in properly running the program?

Not yet but it will be good to compare notes.
I downloaded version BayesCall 0.3, compiled it with GSL 0.13. Run into:

===
Traceback (most recent call last):
File "EM.py", line 557, in <module>
if any(all(array(intensity)<=0.1, 1)):
File "/soft/general/python-2.5.2/lib/python2.5/site-packages/numpy/core/fromnumeric.py",
line 1422, in all
return all(axis, out)
ValueError: axis(=1) out of bounds
===

Got a hint from BayesCall Authors that I should convert my cif files to txt using cif2txt.
Found it here:

http://svncisd.ethz.ch/repos/cisd/deep_sequencing_unit/trunk/source/bash/

edited it a bit, then did the conversion. At this moment I am running since 20+mins:

python EM.py -u 4 --seq_len 36 test_100104/091220_11AB2AABB_out/s_4_0*int.txt

and it produces a bunch of *_int.mat files.

So I will keep you posted how it went. Did you passed that stage and got errors afterwards?

Darek Kedra

PS I have:
Python 2.5.2,
* print numpy.__version__
1.3.0
* print scipy.__version__
0.7.1rc2

on 2.6.23.1-42.fc8 #1 SMP x86_64 GNU/Linux

[lcollado]

Hello,

Were you able to run BayesCall correctly? Also, do you keep using it and what is your experience vs the different SolexaPipeline versions?

Thank you,
Leonardo

[zero]

Hi Leonardo,

I've done some testing of BayesCall using text files. Initially I did a couple of tiles and compared the qc metrics of the output. The data was much better using BayesCall. Saying that, our input data is pretty poor, so there is lot of room for improvement. Original data was GAIIx using Illumina RTA image processing and pipeline 1.5.1. What you will notice is that it takes a while, i.e. the parameter estimation took several hours (depends on the number of tiles you use), but the base calling itself (since v0.3 using naive basecaller) is fairly quick on a Nehalem 8-core. Our data (2x76 PE run) improved from median Phred = 20 up to base 53 to median Phred = 21 up to last base. So I would say its definitely worth a try.

[nilshomer]

Hi Leonardo,

I've done some testing of BayesCall using text files. Initially I did a couple of tiles and compared the qc metrics of the output. The data was much better using BayesCall. Saying that, our input data is pretty poor, so there is lot of room for improvement. Original data was GAIIx using Illumina RTA image processing and pipeline 1.5.1. What you will notice is that it takes a while, i.e. the parameter estimation took several hours (depends on the number of tiles you use), but the base calling itself (since v0.3 using naive basecaller) is fairly quick on a Nehalem 8-core. Our data (2x76 PE run) improved from median Phred = 20 up to base 53 to median Phred = 21 up to last base. So I would say its definitely worth a try.

Is the Phred calculated after alignment or are they reported by BayesCall? I would be interested to know if it improves the base error-rate after alignment.

[zero]

these scores are from BayesCall itself.

[lcollado]

Thank you for your feedback