SEQanswers

SEQanswers (http://seqanswers.com/forums/index.php)
-   Bioinformatics (http://seqanswers.com/forums/forumdisplay.php?f=18)
-   -   Bfast index splitting and appropriate hash width (http://seqanswers.com/forums/showthread.php?t=9813)

MolecularToast 03-02-2011 12:14 PM

Bfast index splitting and appropriate hash width
 
I am creating colorspace reference indexes with bfast and need to split them using the "-d" argument (only 16GB of RAM). Something like this:

$BFAST/bfast index -f $REF -m 1111111111111111111111 -d 1 -w $HASHWIDTH -i $INDEXNUMBER -A 1 -n $CORES -T $TMP_DIR

In the Bfast manual it states that the hash width ("- w") should be reduced from 14 if splitting indexes but I cannot find where it mentions what value should be used.

Page 57: We suggest a hash width of 14, although this should be reduced if you are splitting the indexes for low-memory computation.

I am only planning on splitting into 4 ("-d 1"). Does anyone know what is the most appropriate hash width for this setting? Is there a defined relationship between key size, key width, or another metric and hash width?

I appreciate any assistance.

nilshomer 03-02-2011 11:33 PM

You should be fine keeping the hash width the same. See the BFAST paper's supplemental materials for the full details.


All times are GMT -8. The time now is 10:42 AM.

Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2020, vBulletin Solutions, Inc.